PIIS147149991417301430

Feature Review
Precision Oncology: The

Road Ahead
Daniela Senft,1 Mark D.M. Leiserson,2,6 Eytan Ruppin,3,4 and
Ze’ev A. Ronai1,5,*
Current efforts in precision oncology largely focus on the benefit of genomics-
Trends
guided therapy. Yet, advances in sequencing techniques provide an unprec-
Genomics-driven cancer therapy ben-
edented view of the complex genetic and nongenetic heterogeneity within efits a subset of patients, although
individual tumors. Herein, we outline the benefits of integrating genomic and there are clear shortcomings to this
approach.
transcriptomic analyses for advanced precision oncology. We summarize rel-
evant computational approaches to detect novel drivers and genetic vulner- Using genomics as a single ‘biomar-
abilities, suitable for therapeutic exploration. Clinically relevant platforms to ker’ to inform therapy is insufficient to
comprehensively predict efficient ther-
functionally test predicted drugs/drug combinations for individual patients are apeutic approaches. By providing
reviewed. Finally, we highlight the technological advances in single cell analysis information about active pathways,
the inclusion of transcriptomic data
of tumor specimens. These may ultimately lead to the development of next-
reveals a more comprehensive and,
generation cancer drugs, capable of tackling the hurdles imposed by genetic thus, accurate molecular profile, which
and phenotypic heterogeneity on current anticancer therapies. likely improves the choice of therapy.
Available patient-derived functional

Precision Medicine Aims to Address Inter- and Intratumor Heterogeneity models (e.g., organoids or patient-
Precision medicine aims to use multiple types of data to classify patients into groups that will derived xenografts) are promising for
most likely respond to a given treatment. The identification of biomarkers (see Glossary) that testing multiple drugs and/or drug
combinations in a clinically relevant
correlate with response to therapy or function in disease initiation and/or progression (therefore
time-frame.
representing therapeutic targets themselves) is fundamental in this process [1]. Determination
of molecular biomarkers is not limited to a specific methodology, and DNA, RNA, proteins, Mining available data sets can allow
metabolites, or microorganisms can individually, or in combination, serve as biomarkers. With researchers to comprehensively map
the processes that drive cancer and
cancer primarily being a genetic disease, precision oncology has largely focused on the reveal novel vulnerabilities.
determination of genetic biomarkers and multiple clinical trials test whether targeting these
genetic alterations in cancer can prolong survival. Remarkable success in applying genomics- Intratumor heterogeneity remains one
driven cancer therapy has been noted [2], yet, serious criticism remains regarding this geno- of the biggest challenges in reaching
sustained therapeutic responses to
mics-focused precision oncology concept, including scientific, social, ethical, and economical cancer treatment. Integrating addi-
aspects [2–5]. In this review, we focus on the biological rationale for precision oncology and tional factors (immune, metabolome,
outline current efforts and achievements of implementing precision oncology in the clinic, while and microbiome) could pinpoint novel
putative therapeutic approaches and
highlighting promising routes to overcome the limitations of genomic-focused approaches. The
combinational drug therapies, in an
current availability of screening platforms and the armamentarium of anticancer drugs now effort to overcome tumor
allows us to recognize and address intertumor heterogeneity (i.e., the different molecular heterogeneity.
characteristics observed between patients). We outline how the simultaneous assessment of
genomic and transcriptomic data, combined with functional testing, can serve to overcome
hurdles imposed by intertumor heterogeneity. In addition, we discuss the major limitations of
prolonged response to current anticancer therapies, including intratumor heterogeneity
1
(ITH); namely, differences in the molecular make-up of tumor cells within individual patients. We Tumor Initiation and Maintenance
Program, NCI designated Cancer
have only begun to decipher and address such challenges therapeutically. Center, Sanford Burnham Prebys
Medical Discovery Institute, La Jolla,
The Technical and Molecular Basis for Precision Oncology CA 92037, USA
2
Microsoft Research New England,
The ability to detect mutations in a tumor sample was one of the first milestones in recognizing Cambridge, MA 02142, USA
the genetic events that underlie the cellular transformation process, denoting an early phase of 3
School of Computer Sciences and
874 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 http://dx.doi.org/10.1016/j.molmed.2017.08.003
© 2017 Elsevier Ltd. All rights reserved.
genetic-based evidence for cancer occurrence and development. Improved technologies Sackler School of Medicine, Tel Aviv
University, Tel Aviv, 69978, Israel
enabling the detection of such mutations in non-neoplastic tissues (including bodily fluids) 4
Center for Bioinformatics and
has allowed the early detection of somatic oncogenic mutations, such as Ras mutations and Computational Biology, University of
hotspot p53 tumor suppressor mutations [6–8]. While these developments reflect advances Maryland, College Park, MD 20742,
USA
made already during the 1980s, it has taken another generation to better establish the 5
Technion Integrated Cancer Center,
importance of mutation frequency, its variability in the transformed tissue, and its causative Faculty of Medicine, Technion, Israel
role. This growing understanding has been a prerequisite for the introduction of mechanism- Institute of Technology, Haifa, 31096,
Israel
based therapies into clinical practice. Commonly known as targeted therapies, these 6
Center for Bioinformatics and
therapeutic approaches are based on small molecules or monoclonal antibodies that inhibit Computational Biology, University of
oncogenic drivers [9–14], or target genetic vulnerabilities [e.g., poly (ADP-ribose) polymerase Maryland, College Park, MD 20742,
USA
(PARP) inhibitors in tumors with homologous recombination deficiency [15]]. Several years
of clinical experience with targeted agents, and especially of the resistance to drugs, has led to
the recognition of the central role of genetic heterogeneity and plasticity of growth-promoting *Correspondence:
signaling pathways in determining a patient’s individual response. A notable example is the [email protected] (Z.A. Ronai).
targeting of BRAF mutations, which are present in more than 40% of melanomas [16]. Although
targeting recurrent BRAF mutation(s) by mutant-specific BRAF inhibitors demonstrated great
clinical success [9,17], understanding the complex feedback and crosstalk between key
players of the altered RAS/RAF/MEK/ERK signaling axis became necessary for optimizing
therapy. Accordingly, in terms of clinical outcomes, combined BRAF and MEK inhibition proved
superior over single-agent use [18]. Furthermore, new generations of specific BRAF inhibitors
are currently in the pipeline, finely tuned to overcome mutation-driven altered signaling events in
the RAS/RAF/MEK/ERK pathway [18]; these might be expected to outperform previous
inhibitors of this pathway. Similar undertakings may be required to target deregulated signaling
pathways arising from other mutations in different tumors, where a driver mutation is known,
and where drugs targeting a given driver may exist.
Beyond direct targeting of genomic alterations, the impact of differentiation hierarchies,

epigenetic alterations and the role of the microenvironment in driving tumor pathogenesis
have become increasingly recognized. Accordingly, therapeutic approaches that aim to restore
normal differentiation programs, such as all-trans retinoic acid in acute promyelocytic leukemia
and neuroblastoma, have been developed [19]. Along these lines, drugs are and/or have been
developed to reprogram epigenetic marks and restore normal gene expression programs, such
as various histone deacetylase (HDAC) inhibitors [20], in addition to drugs that interfere with
tumor–microenvironment crosstalk, including angiogenesis inhibitors [21] and immunothera-
peutic agents [22].
The search for cancer vulnerabilities in specific cancer types has been facilitated by numerous
technological advances yielding large-scale molecular profiling of major cancer types [23,24].
This system-based analysis of tumor samples, together with massive hypothesis-based
research, has significantly changed our understanding of cancer biology (Figure 1, Key Figure):
carcinogenesis is generally considered to be driven by the natural selection of continuously
acquired genetic and epigenetic variation in individual cells [25]. These converge on common
phenotypic characteristics for cancer cells, including sustained proliferation, migration, inva-
sion, and/or resistance to apoptosis [26]. Tissue microenvironments provide the fitness
selection defining spatial and temporal changes in environmental pressures. These influence
the evolutionary path of any given cancer cell, resulting in (epi-)genetically heterogeneous
subpopulations. Diversity within cancer cell populations is not limited to the genome, and
dynamic variations in differentiation hierarchies, transcriptional signals, and the proteomic
landscape add to the phenotypic heterogeneity observed within tumors [27]. Indeed, cancer
cells do not exist as isolated entities, but rather, engage in heterotypic interactions with stromal
cells and cooperate with adjacent tumor subclones; this is important, because it can result in
the increased robustness of a tumor [28].
Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 875

Key Figure Glossary
Acquired resistance: (or secondary
Determinants of Tumor Pathogenesis and Measures to Inform Precision resistance) indicates that a tumor
Therapy that initially responded to therapy
becomes resistant to this treatment
during the course of therapy. By
contrast, in intrinsic (primary)
resistance, no responses against the
tumor are noted upon initiation of
therapy.
Actionable mutations: gene
alterations that can be specifically
targeted with an approved or
investigational drug. The term does
not provide information about drug
efficacy.
Afatinib: tyrosine kinase inhibitor of
EGFR (ErbB1), HER2 (ErbB2), and
HER4 (ErbB4).
Angiogenesis: blood vessel
formation.
Basket trial: histology-agnostic trial
design that tests the efficacy of
specific drug(s) in molecularly
stratified patients. It evaluates
whether a biomarker (signature) is
predictive for drug response
irrespective of tumor histology.
Binary alterations: binary
classification of a molecular event,
such as somatic mutations (present
or absent), gene expression
(upregulated or downregulated), or
DNA methylation (hypo- or
hypermethylated).
Biomarker: a molecular
characteristic with a correlative or
functional association with disease
risk, prognosis (prognostic
biomarker), or response to treatment
(predictive biomarker).
Canalization evolutionary
process: describes the stability of a
phenotype despite variation in the
genotype.
Cancer hallmarks: cellular and
molecular functions required for
cancer development and
progression. Hallmarks are
sometimes described by a set of
genes that perform a specific
function.
Cetuximab: anti-EGFR monoclonal
antibody that binds to the
extracellular domain of EGFR and
prevents its dimerization.
Circulating tumor cells (CTCs):
tumor cells that can be found in, and
isolated from, the circulation of blood
and/or lymphatic system of patients
with cancer.
Circulating tumor DNA (ctDNA):
circulating, cell-free tumor DNA that
(See figure legend on the bottom of the next page.)
876 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Moreover, large-scale sequencing of human cancer genomes and transcriptomes have identi- can be isolated from whole blood of
fied nearly 200 ‘consensus’ driver genes (of which 15% were identified primarily using DNA patients with cancer.
Clone: one or more cells derived
sequencing of cancer genomes [29]) and an additional 300 putative driver genes have been from, and genetically identical to, a
suggested [30,31]. The pathways in which these genes function are also emerging [32–36]. single ancestor cell. Accordingly,
Coupled with the success seen using targeted therapies in certain cancer subtypes [9–14], subclones share many of the genetic
these efforts laid the basis for precision molecular oncology: patients are treated according to features of the initial ancestor cell,
but contain additional genetic
the molecular make-up of their tumors rather than solely based on tumor histology, type, grade, alterations.
and stage (Figure 1). Conditional reprogramming:
technique used to establish patient-
Clinically Relevant ‘Omics Approaches derived cell cultures from healthy or
diseased (e.g., tumor) tissue.
Genomics-Driven Cancer Therapy in Clinical Testing Deep neural nets: neural network
While at present only a small proportion of cancer patients benefit from targeted therapies, of multiple layers often used for
great efforts are ongoing to extend the scope of precision oncology to a broader spectrum of supervised learning; at each layer, a
function is applied to the input from
patients (reviewed in [2]). Massive intertumor heterogeneity has been rigorously documented
the previous layer.
through large-scale DNA and RNA sequencing, as well as DNA copy number profiling and Differentiation hierarchies: relates
DNA methylation profiling [e.g., The Cancer Genome Atlas(TCGA), International Can- to differences in the impact of
cer Genome Consortium(ICGC), and others] [23,24] (see Table S1 in the supplemental phenotypes seen in an isogenic
population of cancer cells, such as in
information online for relevant data sources). However, unexpected similarities between tumors their ability to metastasize or form
of different tissues of origin have been uncovered, while certain tumors have been found to be tumors upon serial transplantation on
more similar at the molecular level to tumors from a different tissue of origin [37]. These immunodeficient mice (i.e., cancer
similarities, together with the detection of rare variants within well-characterized driver genes, stem cells).
Dimension reduction: selecting a
suggest that approved targeted therapies might be effective in diverse tumor types with distinct subset of features, or combining
molecular alterations [32,36]. This has resulted in the initiation of clinical programs that evaluate features, from a data set (e.g.,
whether molecular profiling of patients is clinically feasible and, importantly, whether treating principal component analysis).
DNA copy number profiling: the
patients based on their genomic profiles might be beneficial relative to a given standard of care,
genome-wide screening for gene
or a physician’s treatment choice (see Table 1 for examples of programs and/or studies). In copy number variations (gains or
addition, the identification of new putative driver genes found in a low percentage of patients losses).
with less-common cancer types or subtypes has generated several novel clinical hypotheses DNA methylation profiling:
genome-wide screening for variations
that await verification. Recently, Foundation Medicine reported that, in a targeted sequencing
in DNA methylation status (hyper- or
study of 63 220 tumors, more than 75% of patients presented a mutation in at least one of ten hypomethylation).
cancer driver genes, and more than 25% of patients presented a known driver mutation within Driver: usually refers to a genetic
these genes [38]. Accordingly, in silico computational studies predict that up to 90% of patients event that is shown to elicit
phenotypic changes associated with
may benefit from molecularly guided therapy when biomarkers of uncertain clinical significance,
tumor initiation and progression (see
as well as off-label and investigational drugs are considered to inform therapy [39,40]. To test ‘oncogenic drivers’). In a broader
this multitude of novel clinical hypotheses, new adaptive trial designs, including basket and definition, the term can also be used
umbrella trials, have been used [41,42] (Table 2). Basket trials are designed to test the effects to describe nongenetic and/or non-
cell autonomous alterations that can
of a single (or a few) drug(s) in a variety of cancer types (or possibly subtypes) using specific alter certain aspects of disease
mutation(s) as biomarker(s). By contrast, umbrella trials are designed to test the impact of progression.
specific drugs on different mutations within the same cancer type. Epigenetic alterations: heritable
trait not explained by changes in
DNA sequence but by changes in
gene expression. Common examples
Figure 1. The genetic and phenotypic characteristics of a patient’s tumor are influenced by tissue- and/or cell type- often observed in cancer cells
specific factors, germline genetic background, lifestyle factors, as well as the number and type of previous anticancer include promoter hypermethylation or
drugs received [25,27,28]. Each individual cell is further influenced by, first, the proximity to and the integrity of the tumor aberrant histone modifications (e.g.,
vasculature; second, the biochemical and biophysical properties of the surrounding extracellular matrix (ECM); third, acetylation).
competing and/or cooperating interactions between individual tumor cells or tumor and stromal cells [among which are Evolutionary pressure: (or selection
cancer-associated fibroblasts (CAFs), endothelial cells (ECs), and bone marrow-derived cells (BM-DCs)]; and fourth, pressure) any change in the
antitumor immunity. These factors further shape the geno- and phenotypic properties of a tumor in a spatial and temporal microenvironment (e.g., cancer
manner. While the genomic analysis of a tumor biopsy at the time of diagnosis identifies genetic vulnerabilities, the inclusion therapies) that leads to a selection of
of transcriptomic data holds the additional potential of identifying nongenetic vulnerabilities by considering pathway activity clones that have a growth advantage
and the composition of the tumor microenvironment. Therefore, the integrated analysis of genomic and transcriptomic under these new conditions.
data is a valuable tool to inform precision therapy. Abbreviations: seq, sequencing; UV, ultraviolet. Functional mutations: mutations
that change the phenotype of a
cancer cell or tumor.

The majority of these studies profile the mutation status of a few dozen or hundreds of selected Gene fusions: hybrid genes that
genes [2]. This is based on the fact that, although whole-genome sequencing (WGS) can detect combine parts of two or more
original genes. Fusion genes
DNA sequence variants as well as focal and large chromosomal rearrangements, deletions, or originate from chromosomal
amplifications, it is difficult to identify driver events within large chromosomal abnormalities. rearrangements (i.e., deletions,
Therefore, clinically valuable sequencing approaches can be reduced to either the whole translocations, tandem duplications,
exome (WES) or targeted exomes (panel sequencing) of cancer-related genes (Figure 1). or inversions).
Genetic interaction: functional
These approaches are often combined with the analysis of some well-characterized intronic interplay between multiple genes and
regions (e.g., ALK, RET1, ROS, or BCR) that are frequently rearranged in cancer genomes [2]. their corresponding gene products
Furthermore, targeted sequencing has the advantage of yielding a high sequencing depth, that impacts the cellular phenotype.
Genotype-matched trials: when a
which is important to be able to infer the clonality of a detected driver event. Determining the
clinical trial is selected for a patient
clonal distribution of identified alterations should be a priority in precision oncology trials, given based on their genotype. This can
that targeting trunk mutations appears to be crucial to maximizing the efficacy of targeted be a study that only accepts patients
therapies [43]. with a given mutation, but also any
study that either analyses a drug
likely to be effective in the context of
Most studies demonstrated that genomics-guided therapy improves patient outcomes when a given genotype, or inhibits a
well-characterized biomarker–drug pairs, supported by strong clinical or preclinical evidence, pathway that is directly linked to a
are used (Table 1). For example, the MOSCATO 01 trial [44], which used only last-generation mutation identified in the patient can
be considered genotype matched.
drugs with high affinity to a specific target, achieved positive results, whereas the SHIVA trial Histone deacetylase (HDAC)
[45], which heavily relied on everolimus, a drug weakly affecting the PI3K/AKT/mTOR pathway, inhibitors: HDACs remove acetyl
indicated that genomic profiling did not result in patient benefit. Furthermore, emerging groups from histone lysine residues,
evidence suggests that targeting multiple drivers by combination therapy is superior over but also from nonhistone substrates.
HDAC inhibitors can have not only
single-agent use [46], because patients with advanced tumors frequently exhibit multiple transcription-dependent effects (e.g.,
aberrations detected by genomic profiling. Despite these advances, several challenges remain. relief of transcriptional repression of
First, trial recruitment of patients with rare mutations is difficult [47] and only a small percentage tumor suppressor genes), but also
transcription-independent effects due
of patients (2–5%) undergoing genomic profiling have subsequently been treated with off-label
to altered acetylation (and activity) of
drugs or been enrolled in genotype-matched trials [46,48]. Noteworthy, the recent report on nonhistone substrates, such as those
the positive outcomes of the MOSCATO 01 trial indicates that match rate can be improved involved in cell proliferation or cell
(19%) when performed within a cancer center offering access to a variety of clinical trials [44]. death.
Homologous recombination
Second, the presence of validated genetic biomarkers does not strictly predict a response to
deficiency: a defect in double-
targeted therapies in different tumor types [49], given that the effect of therapies is context strand break repair by homologous
dependent [as seen, e.g., by the lack of response of BRAFV600-positive colorectal cancer (CRC) recombination repair associated with
to BRAF inhibitors, which show good clinical responses in melanomas carrying the same high levels of genomic instability.
These defects were originally
mutation [50]. On a positive note, identifying such genomic-context effects has already
identified in BRCA1/2-deficient
expanded the use of inhibitors, such as against BRAF [49] or PARP [51], and may result in tumors, but can also be observed in
the expedited approval of investigational drugs. Finally, identifying high-confidence biomarkers the absence of BRCA mutations, a
to guide specific drug treatments remains a challenge. One approach to expanding the phenomenon generally referred to as
‘BRCA-ness’.
biomarker landscape may be to determine the differential molecular profile of patients showing
ICGC-TCGA DREAM Somatic
a dramatic response to targeted therapy versus nonresponsive patients. An increasing number Mutation Calling Challenge –
of publications report such ‘exceptional responders’, which has led to the identification of rare RNA: the DREAM Challenges are
genomic events likely to predict the response or resistance to targeted therapies [52–62]. crowdsourcing (open science efforts)
challenges examining questions in
Taken together, sequencing efforts of cancer genomes within clinical trials or by research biology and medicine. In the ‘RNA-
initiatives, such as the TCGA and ICGC initiatives (Table S1 in the supplemental information Challenge’, leaders from ICGC and
online) are expected to improve the identification of driver mutations, as well as patient TCGA have come together to
develop a Challenge to assess the
stratification strategies associated with these. This in turn may expand the scope of geno-
accuracy of methods to work with
mics-based precision oncology to a broader spectrum of patients. cancer RNA-seq data.
Immune-checkpoint inhibitors:
Limitations of Using Genomics As a Single Approach for Biomarker Identification drugs that block immune-inhibitory
signals (such as PD1, PD-L1, or
While genomic profiling provides valuable information regarding genetic mutation, amplifica-
CTLA4) expressed on and/or by
tion, deletion, and certain epigenetic modifications (e.g., methylation), there are certain inherent tumor or immune cells. Inhibition of
limitations of using an approach that simply tests the presence or absence of genetic driver these factors can unleash, in some
events to inform therapeutic decision-making. This includes limitations of using genomics as a cases, a robust and durable
antitumor immune response.
single platform for biomarker identification; indeed, in some cancer types, such as prostate

cancer or some pediatric malignancies, few or even no recurrent mutations have been ‘Integrated’ subtype: a subtype of
detected, indicating that other types of somatic variation may be potent drivers of oncogen- cancer that includes patients with
tumors from multiple tissues of
esis [36]. Furthermore, no genetic alterations have been found to correlate with well-charac- origin.
terized predictive biomarkers, such as the expression of estrogen receptor or androgen International Cancer Genome
receptor in breast or prostate cancer, respectively. In addition, genomic profiling does not Consortium (ICGC): has the goal to
provide sufficient information regarding the activity of actual protein products mediating generate comprehensive catalogs of
genomic abnormalities (somatic
oncogenic or tumor suppressor gene functions. In other words, variations in oncogenes mutations, abnormal expression of
and/or tumor suppressor genes do not necessarily predict activation of the corresponding genes, and epigenetic modifications)
biological pathway, and vice versa: cancer driver pathways can be active without the presence in tumors from 50 different cancer
types and/or subtypes and make the
of a mutation(s) [63]. Finally, novel biomarkers linked to nongenetic vulnerabilities, such as those
data available to the entire research
involving cancer cell reliance on stress response or metabolic pathways, that may be able to community to accelerate research
predict responses to autophagy inhibitors or drugs inhibiting antioxidant enzymes need to be into the causes and control of
defined [64]. cancer.
Intertumor heterogeneity:
differences in the molecular make-up
The most comprehensive approach to overcome these challenges and to elucidate cancer of tumor cells between patients.
vulnerabilities is the simultaneous characterization of the genome, epigenome, transcriptome, Intratumor heterogeneity (ITH):
proteome, and metabolome of tumors and their surrounding stroma; indeed, these are all differences in the molecular make-up
of tumor cells within individual
crucial parameters to defining cellular phenotypes involved in cancer pathogenesis, as well as in patients.
characterizing responsiveness to therapy [65] (Box 1). Given that these parameters are Match rate: frequency by which
dynamic entities (e.g., change in responses to external stimuli), they are expected to show genomic alterations detected in a
spatial heterogeneity (geno- or phenotypic distinct clones may show different growth kinetics patient cohort can be matched to
targeted therapies.
or survival rates dependent on their location). Consequently, an analysis of multiple biopsies Methylome: collection of all DNA
methylation markers in a single cell
or a population of cells.
Box 1. Relevant ‘Omics for Precision Oncology MOSCATO 01: prospective clinical
Epigenetic profiling holds great promise in deciphering cellular states and characterizing phenotypic heterogeneity. The trial evaluating whether high-
importance of epigenetic reprogramming in cancer is evidenced by the fact that chromatin regulators are often mutated throughput genomic analyses can
[32,36] and widespread epigenetic changes throughout cancer genomes can be identified, intricately linked to the improve patient outcomes.
activity of known tumor promoters and/or suppressors, such as EGFR [200] or TP53 [201]. There are two general ‘Multiview’ matrix factorization:
classes of drugs targeting the epigenome: (i) broad reprogrammers, which include inhibitors of DNA methyltransferases, the ability to factor multiple data
histone deacetylases, or bromodomain and extraterminal motif proteins; and (ii) targeted therapies that pin specific matrices into a lower-dimensional
activating mutations in DNA-modifying enzymes, such as EZH2, or in enzymes whose mutations have a profound effect space, where each data matrix gives
on epigenetic pathways, such as IDH1/2 [20]. Currently, there are no epigenetic drug-sensitivity biomarkers that would a different ‘view’ of the data (e.g.,
predict the response to these approved or investigational drugs. Therefore, the addition of epigenetics in clinical practice gene expression and DNA
awaits the identification of epigenetic marks that mediate distinct tumor phenotypes of clinical relevance (such as methylation views).
mesenchymal differentiation, stemness, dormancy, or therapy resistance) [65]. Off-label (drug use): a drug is used
for a purpose not specified in the
marketing authorization determined
Proteomics combined with genomic data likely reveal the most accurate information on the activity state of individual
by a licensing body, such as the FDA
genes. The proteome represents the ideal readout to define the functional state of a cell in response to internal or
(e.g., drug used for a different type
external perturbations, and proteogenomic analysis is being integrated in large-scale characterization efforts of the
of cancer than the one it is approved
TCGA [202–204]. This integration has the power to nominate driver genes from large chromosomal deletions or
to treat).
amplifications and can identify new driver clusters that are not easily found in transcriptomics signatures [202,203,205].
Oncogenic drivers: genetic events
Although TCGA analysis has long included antibody-based phosphoprotein analyses, the comprehensive proteomic
associated with tumor initiation and
characterization based on mass spectrometry increases the breadth of phosphoproteomics data and, importantly,
progression, including metastasis
allows for the identification of post-translational modifications beyond phosphorylation [206]. The latter may represent
and therapy resistance.
important biomarkers for drugs that do not target kinases, such as the identification of ‘acetylation signatures’ in serous
Orthotopic implantation: a form of
ovarian cancer, and which may predict responses to HDAC inhibitors [204]. While it is expected that future technologies
xenograft experiment where (patient-
will provide the platform for large-scale proteomic assessment of tumor samples, current proteomic analysis requires a
derived) tumor cells are implanted
large amount of tissue, is costly, labor-intensive, and lacks the analytical validity and sensitivity that genomics provides.
into the organ of origin to maintain
the tissue-specific environment.
Emerging metabolome and microbiome data are expected to provide important additions to genomics: Rewired Paclitaxel: chemotherapeutic drug
metabolic pathways in tumors provide alternate fuel sources that can be targeted, and the mutations and/or that binds tubulin and inhibits the
deregulated expression of metabolic genes have been linked to tumor propensity for metastasis or therapeutic disassembly of microtubules, thereby
resistance [207]. Translating this knowledge in the clinic will require further preclinical analysis, especially given the resulting in the inhibition of cell
differences between cancer cell metabolism in vitro and in vivo [208,209]. Microbiome-based data are a likely addition in division.
the more distant future, which might provide novel putative biomarkers and means to monitor predicted therapeutic Panitumumab: monoclonal antibody
responses and, possibly, lead to the development of new therapeutic modalities. that binds to and inhibits (autocrine)
stimulation of the EGFR.

and longitudinal follow-up of patients would ideally be performed to predict the initial responses Passenger mutations: somatic
to therapy and to identify putative mechanisms of drug resistance. Although such compre- mutations with no obvious role
(owing to their inability to cause
hensive approaches are not yet feasible for routine clinical practice, current state-of-the-art phenotypic changes) in cancer.
technologies are already enabling the combination of at least two different ‘omics platforms for Patient-derived xenograft (PDX):
cancer analysis, genomics with epigenomics, and/or transcriptomics. As discussed below, xenograft model where tumor cells
combined genomic and transcriptomic analysis, together with functional testing of ‘omics- (derived from a biopsy or bodily
fluids) are subcutaneously or
derived treatment predictions, are expected to overcome many of the challenges that current orthotopically engrafted into immune-
precision oncology-based trials face. incompetent mice, without any
intermediate growth or modification
in culture.
Transcriptomics As a Valuable Measure to Improve Biomarker Identification
Phenotypic heterogeneity:
At present, the most common way to enhance genomic information is by the inclusion of differences in phenotypic
transcriptomic analyses (Figure 1). RNA sequencing (RNA-seq) technologies allow the mapping characteristics (e.g., potential to form
of the entire transcriptome or select gene expression networks, and are readily available and metastasis or resist therapy) of
genetically diverse but also isogenic
becoming economically feasible. The ability to decipher the landscape of gene expression
tumor cells.
offers important steps over acquiring genomic data alone. First, aside from the ability of RNA- Rare variants: genetic variations (i.
seq to detect splice variants [66], it can also detect novel or known gene fusions, which have e., mutations) within a particular gene
been identified as drivers of disease in a variety of rare and common tumors [66–69]. Gene that occur in <1% of patients.
Recurrent mutation: somatic
fusions are promising therapeutic targets because the inhibition of fusion genes is often mutation that occurs in a statistically
associated with striking efficacy, as exemplified by targeting the BCR-ABL fusion in chronic significant number of times in a
myeloid leukemia, or targeting RET, ALK, ROS, FGFR, or BRAF fusions in various malignancies cohort of sequenced tumors.
(reviewed in [2]). Second, transcriptomics can provide indirect information about protein Sample-level events:
measurements of individual tumors,
expression status; knowing that a candidate gene harboring certain mutation(s) is also patients, or cancer cell lines (e.g.,
expressed (and to what level) is valuable in establishing the importance and contribution of drug sensitivity).
this gene to the tumor phenotype. Third, beyond providing information about the expression of Saturation analysis: modeling of
sample complexity to determine the
tumor driver genomic variations, the inclusion of transcriptomics allows the mapping of non-
number of samples required to
oncogene vulnerabilities, and provides information about oncogenic pathway activities, even in detect some event as statistically
the absence of mutated driver genes [63,70]. One such example is the BRAF mutation significant.
signature in colon cancer, which can be found in not only BRAF mutated but also BRAF Sequencing depth: reflects the
average number of times a given
wild-type tumors, and characterizes (in addition to the KRAS and PI3K signatures) patients
region has been sequenced by
resistant to EGFR inhibition [71]. The BRAF signature can not only serve as a resistance independent reads.
biomarker, but has also been recently suggested to serve as a sensitivity biomarker for mitotic Serial passaging: serial
poison drugs, such as vinorelbine [72]. Another notable example is that of BRCA-associated transplantation of tumors (usually in
the flank of nude mice) for tumor
signatures [73], where tumors (such as breast, ovarian, or prostate tumors, among others
expansion.
[66,73]) sharing similar molecular signatures to BRCA mutant tumors may also respond to SHIVA study: the first randomized,
similar therapeutic approaches, even when lacking specific BRCA mutations [73–75]. Fourth, open-label, controlled Phase II study
transcriptomes, by contrast to DNA, are tissue and cell type specific [65]; this is often that evaluated whether targeted
treatment based on molecular testing
considered a disadvantage for RNA analysis of bulk tumor samples because, in samples with improved patient outcomes
a high proportion of stromal cells, massive computational deconvolution is necessary to extract compared with conventional
the transcriptional profile of interest, given that all cells within the biopsy contribute to the RNA treatments across all tumor types.
pool [65]. However, tissue specificity can provide important clinical information about tumor Somatic variation: alteration
(genetic, epigenetic, or signaling)
histology and tumor origin, which is of relevance in patients with cancer of unknown primary occurring in a somatic (body) cell.
tumors [37]. Additionally, cell type-specific transcriptomes can reveal certain aspects of the Splice variants: transcripts (mRNAs)
immune status of tumor samples that may be of therapeutic relevance [76,77]. High overall resulting from the alternative splicing
of different exons in genes.
mutational load within tumors (e.g., highly mutated human tumors, such as melanoma, lung
Suppression interactions:
cancer, or mismatch-repair deficient colon cancer) has been reported to correlate with phenotypic defects caused by a
therapeutic responses to immune-checkpoint inhibitors (e.g., drugs targeting CTLA-4 or mutation in a particular gene are
the PD1/PDL1 axis) [76,77]; however, these factors have not been strictly linked, and long-term rescued by a mutation and/or
deregulation in a second gene.
responses to checkpoint inhibition have been observed for a broad mutational spectrum of
Synthetic rescue: type of genetic
cancers [76,77]. Integrating genomic and transcriptomic data holds promise for the identifica- interaction where the mutation and/
tion of patients who can benefit from immune-checkpoint inhibitors. For example, the tran- or deletion of one gene rescues the
scriptomic analysis of responders to CTLA-4-blockade (ipilimumab) revealed that the lethality or growth defect of a cell
expression of cytolytic effector genes (e.g., those encoding granzyme A and perforin) positively

correlate with patient response (complete or partial response to ipilimumab, or stable disease mutated and/or deleted for another
with overall survival >1 year by RECIST criteria) [76]. Furthermore, the expression of immune gene.
Targeted therapies: drugs that
checkpoint regulators has correlated with increased patient survival [76]. In addition, tran- either target molecular alterations
scriptomic signatures that significantly correlate with resistance to anti-PD1 therapy in mela- specific to cancer cells (e.g.,
noma have also been identified [77]. mutated, amplified or epigenetically
up- and/or downregulated signaling
proteins), or target immune cells to
Although the practical utility of RNA-seq in the clinic has been challenging, technological increase anticancer immunity.
advances allowing the application of RNA-seq to clinically relevant specimens (including Temozolomide: chemotherapeutic
formalin-fixed, paraffin-embedded tissues), along with efforts to benchmark data analysis drug in the class of alkylating agents
exerting cytotoxicity by inhibiting
pipelines (ICGC-TCGA DREAM Somatic Mutation Calling Challenge – RNA) [78], have
DNA replication.
set the basis to move RNA-seq into routine clinical practice. Thus, valuable transcriptomic The Cancer Genome Atlas
information can be combined with genomic data to establish new blueprints that provide (TCGA): a collaboration between the
multidimensional insight into the characteristics of a given tumor biopsy. Such combinations National Cancer Institute (NCI) and
the National Human Genome
can benefit from innovative computational approaches that may identify novel master regu-
Research Institute (NHGRI); it has
lators, not seen in either analysis alone. generated comprehensive,
multidimensional maps of key
Analysis Approaches to Determine Molecular Subtypes and Cancer genomic changes in 33 types of
cancer.
Vulnerabilities Trametinib: targeted therapy that
To overcome the challenges of intertumor heterogeneity in determining molecular-guided specifically binds and inhibits MEK1/
therapy, the identification and characterization of molecular subtypes of cancer and the 2.
mutations that drive cancer have been an urgent priority. The promise of characterization Trunk mutations: genetic variations
occurring in early tumor evolution
of tumors with molecular subtypes or biomarkers is twofold. The first major goal is to find and are thus common to most
molecular biomarkers of patient prognosis or effective drug treatments. The second major goal clones.
is to develop a better mechanistic model for understanding the role of the genome, tran- Umbrella trial: study design
evaluating the efficacy of multiple
scriptome, methylome, epigenome, and environmental alterations of the tumor in driving its
drugs within a biomarker-stratified
initiation and evolution. Extensive clinical efforts now provide an unprecedented view of the single cancer type. It allows the
genomic (and transcriptomic) landscape of all advanced cancer types [79], in addition to the evaluation of multiple biomarker–drug
data sets provided by the TCGA and ICGC, which have focused on the sequencing of common combinations in a histology-
dependent manner.
cancer types early in disease progression.
The following sections describe related computational approaches; for an expanded summary
of references on these and additional topics (ITH and single cell analysis approaches) the reader
is referred to Tables S1–S3 in the supplemental information online.
Approaches for Tumor Subclassification

Methods for identifying molecular subtypes generally fall into two categories, based on whether
data from a single platform or multiple platforms are being used. For single-platform data (e.g.,
gene expression), any off-the-shelf clustering algorithms can be used, although choosing the
method depends on the type of data being clustered. The more challenging case is clustering
patients with data from multiple platforms, especially because there is often a data type that is
missing, given that not all measurements are performed in every patient. Researchers have
taken multiple approaches (see references in Table S2A in the supplemental information online).
Some methods search for a ‘consensus’ after clustering patients by each platform separately
[37], or cluster with protein–protein interactions [29], or patient similarity networks [80,81].
Other methods formulate the problem as a ‘multiview’ matrix factorization and dimension
reduction (reviewed in [82]), or as a probabilistic model. In all cases, a key challenge is the
selection of features from each platform as inputs to the clustering algorithms; for example, it is
possible to summarize mutations, gene expression, and DNA methylation events as binary
alterations [80], and then treat any missing data as a nonalteration event. We anticipate that
recent advances in methods for learning low-dimensional representations of multiple data
types, such as deep neural nets [83], will soon be applied in molecular classification of tumors,
given the amount of molecular cancer data being produced and the successful application of

Table 1. Studies Evaluating Feasibility and Clinical Benefit of Molecular Profilinga
Study Tumor type Screening platform #Pts #Pts on matched Type of therapy Endpoint/Results
sequenced therapy
IMPACT; M.D. Anderson All; advanced, Hotspot sequencing 1144 175 Mono and Higher ORR and longer
[193]; NCT00851032; refractory 11 genes; FISH combination; Phase PFS compared with
retrospective, (ALK) I therapies; N.S. unmatched therapy
nonrandomized
IMPACT/COMPACT [48]; Advanced MALDI-TOF MS 1640 245 N.S.; investigational Genotyped matched
NCT01505400; cancers and hotspot panel (23 agents of 277 trials, therapy improved
retrospective, Phase I genes) or targeted including 89 response
observational, candidates NGS panel (48 or 50 genotype-matched
nonrandomized genes) trials
PREDICT [194]; All NGS; 347 87 N.S. More patients with SD

NCT02478931 (182 or 236 gene 6 months; 45% with
retrospective, correlative, panel and 14 or 19 extended PFS of 30%
nonrandomized rearrangements compared with previous
therapy
Bisgrove [195]; All; advanced, IHC, FISH, gene 86 68 Mono and 27% with extended PFS
NCT00530192; refractory expression combination; N.S. of 30% compared with
prospective, single-arm previous therapy
Phase I
Genomic Profiling in Phase All; advanced, Panel NGS (236 339 122 Mono and High matching score
I [46]; NCT02437617; refractory genes); standard combination; Phase associated with higher
prospective, biomarker I/II therapies; N.S. SD 6 months/PR/CR,
nonrandomized longer PFS and survival
MOSCATO 01 [44]; All; advanced, CGH array; panel 843 199 Phase I drugs and 33% with extended PFS
NCT01566019 refractory NGS (WES and off-label drugs; N.S. of 30% compared with
prospective, RNAseq in 2014 previous therapy
nonrandomized included)
WinTher; NCT01856296; All; advanced, DNA (236 genes) To be 200 N.A. N.S. Chosen based Estimated completion in
prospective, refractory and RNA in tumor on DNA analysis, if 2018
nonrandomized and normal no actionable
matched tissue mutation, then
based on RNA
analysis
SHIVA [45]; NCT01771458 All; advanced, Panel hotspot NGS 741 99 Erlotinib, lapatinib Median PFS in
prospective, randomized refractory (46 genes); CNV + trastuzumab, experimental group
sorafenib, imatinib, (matched therapy) not
dasatinib, significantly longer than
vemurafenib, in control group
everolimus
NCI-MPACT; All, advanced NGS (4000 variants To be 700 N.A. Carboplatin, Estimated completion in
NCT01827384; solid tumors across 143 genes); everolimus, 2019; outcome
prospective, Phase II activation of RAS/ temozolomide, measures: ORR;
randomized feasibility RAF or PI3K trametinib, veliparib, compare genotype
study pathway; AZD1775 matched versus
inactivation of DNA physicians choice
repair pathway
a
Abbreviations: AZD1775, WEE1 inhibitor; CGH, comparative genomic hybridization; CR, complete response; FISH, fluorescent in situ hybridization; IHC, immu-
nohistochemistry; N.A., not applicable; N.S., not specified; NGS, next-generation sequencing; ORR, overall response rate; PFS, progression-free survival; PR, partial
response; Pts, patients; SD, stable disease; WES, whole-exome sequencing.
deep neural nets in areas of computer vision, natural language processing, and biology [84].
Initial molecular subtype studies have often focused on clustering samples into subtypes based
on gene expression in a single cancer type, which have provided robust biomarkers and
subgroups, coherent with patient survival profiles (e.g, in breast cancer [85] or CRC [86]). These
studies typically reveal a more refined set of clusters than those defined by known

Table 2. Representative Basket and Umbrella Trialsa
Study Tumor type Screening platform Biomarkers tested Drugs Endpoint/Results
BATTLE; Chemorefractory Non-NGS, mutation Mutation/amplification: Erlotinib, sorafenib, Better DCR for EGFR
NCT00409968, NSCLC analysis, FISH, IHC EGFR, KRAS/BRAF, vandetanib, erlotinib + erlotinib and KRAS/
NCT00411671, CCND1; protein + bexarotene BRAF +sorafenib
NCT00411632, expression: VEGF, RXR,
NCT00410059, cyclinD1
NCT00410189;
Prospective, adaptively
randomized, umbrella
trial
BATTLE-2; Advanced NSCLC NGS; DNA, mRNA, IHC: pAKT, PTEN, HIF1a, Erlotinib (ctrl) sorafenib, 8-week DCR
NCT01248247; 2- RPPA, IHC LKB1. Mutation: P13KCA, MK-2206 + erlotinib; MK-
stage Phase II umbrella BRAF, AKT1, HRAS, 2206 + selumetinib
design NRAS, MEK1, MET,
CTNNB1, LKB1;
Lung-MAP; Advanced, recurrent FMI FoundationOne PI3KCA, CDK4/6, CCND1/ Taselisib, palbociclib, PFS, ORR, OS
NCT02154490 squamous cell lung platform; IHC; 2/3, FGFR1/2/3, HGF/c- AZD4547, erlotinib,
randomized phase II/III carcinoma MET erlotinib + rilotumumab,
umbrella design ipilimumab, nivolumab,
durvalumab
I-SPY-2; Stage III breast cancer Conventional, MRI HER2, hormone receptor, HER2+: neratinib, MK2206 Pathological complete
NCT01042379; and Mammaprint + trastuzumab, T-DM1 response; several drugs
randomized open-label Bayesian marker-adaptive + pertuzumab, trebananib graduated to Phase III
Phase II umbrella trial designs + trastuzumab, trials
design; (adjuvant pertuzumab [196–198]
setting) + trastuzumab; HER2 :
veliparib, MK2206,
ganitumab + metformin,
trebananib
ALCHEMIST; Resectable NGS, tissue and EGFR mutation; ALK Erlotinib, crizotinib, OS
NCT02194738; nonsquamous NSCLC blood; germline rearrangements (FISH); nivolumab
NCT02193282; + somatic PD-L1 expression (IHC)
NCT02201992; alterations
NCT02595944;
screening study and
accrual to Phase III
randomized treatment
studies (adjuvant
setting)
SAFIR02 Breast; HER2 recurrent and/ CGH array, hotspot To be determined during AZD2014, AZD4547, PFS compared with
NCT02299999; Phase or metastatic breast sequencing study AZD5363, sapitinib, standard maintenance
II randomized umbrella cancer selumetinib, vandetanib,
design bicalutamide, olaparib,
durvalumab
SAFIR02 Lung; EGFR and ALK WT CGH array, hotspot To be determined during AZD2014, AZD4547, PFS compared with
NCT02117167; Phase recurrent and/or sequencing study AZD5363, sapitinib, standard maintenance
II randomized umbrella metastatic NSCLC selumetinib, vandetanib,
design durvalumab
Lung MATRIX; Advanced, pretreated 28-gene NGS Mutation: FGFR2/3, TSC1/ AZD4547, AZD2014, ORR, PFS
NCT02664935 Phase NSCLC platform 2, LKB1, KRAS + RbWT, palbociclib, crizotinib,
II; nonrandomized NF1, NRAS, PIK3CA, selumetinib, AZD5363,
umbrella design AKT1, EGFR + EGFRT790; osimertinib, durvalumab
LoF: p16 + RbWT, PTEN;
amplification: CDK4
+ RbWT, CCDN1 + RbWT,
MET, PIK3CA; rearranged:
ROS1
Hotspot-seq BRAF, PIK3CA, KRAS, ORR, PFS, OS

biomarker only; IHC NRAS; PTEN, MMR

Table 2. (continued)
Study Tumor type Screening platform Biomarkers tested Drugs Endpoint/Results
FOCUS 4 [199]; Phase Advanced/metastatic, BRAFi + panitumumab +/–

II/III randomized untreated colorectal MEKi, aspirin, AKTi
umbrella design cancer + MEKi, HER1/2/3i
V-BASKET [49]; Solid tumors, multiple Mutation analysis BRAFV600 Vemurafenib monotherapy; Efficacy in NSCLC,
NCT01524978; flexible, myeloma with local method vemurafenib + cetuximab ECD, and LCH
early Phase II; basket in CRC
study
CUSTOM [47]; NSCLC, SCLC, thymic NGS Mutation: AKT1, BRAF, Erlotinib, selumetinib, Targeting EGFR and
NCT01306045; malignancy EGFR, ERBB2, HRAS, KIT, MK2206, lapatinib, ALK offers benefit;
biomarker-derived, KRAS, NRAS, PDGFRA, sunitinib design not feasible for
multiarm, multihistology PIK3CA, PTEN; most arms
Phase II, basket trial amplification: ERBB2,
PIK3CA, PDGFRA; fusion:
ALK
NCI-MATCH; Advanced, recurrent, NGS (4000 variants Mutations: AKT1, BRAF Ado-trastuzumab ORR
NCT02465060; refractory solid tumors, across 143 genes) V600; BRAF non-V600, emtansine, afatinib,
nonrandomized Phase lymphoma, myeloma BRCA1/2, cKIT, DDR2, AZD4547, AZD5363,
II; basket trial dMMR, EGFR, AZD1775, osimertinib,
EGFRT790 M, FGFR1/2/3, binimetinib, crizotinib,
GNAQ/GNA11, HER2, dabrafenib + trametinib,
MET ex14 sk, mTOR, NF1, dasatinib, defactinib,
NRAS, PI3KCA, PTEN, GSK2636771, larotectinib,
SMO/PTCH1, TSC1/2; nivolumab, palbociclib,
Translocations: ROS1, sunitinib, TAK-228,
ALK; amplification: taselisib, trametinib,
CCDN1/2/3, CDK4/6, trastuzumab, vismodegib
HER2, MET; loss: NF2,
PTEN; fusion: NFRK
a
Abbreviations: AZD4547, FGFR inhibitor; AZD5363, AKT1/2/3 inhibitor; AZD1775, WEE1 inhibitor; AZD2014, inhibitor of mTORC1/2; CGH, comparative genomic
hybridization; DCR, disease control rate; ECD, Erdheim–Chester disease; GSK2636771, inhibitor of PI3K beta; LCH, Langerhans’ cell histiocytosis; MK2206, inhibitor
of AKT1/2/3; NGS, next-generation sequencing; NSCLC, non-small cell lung cancer; ORR, overall response rate; OS, overall survival; PFS, progression-free survival;
RPPA, reverse-phase protein array; SCLC, small cell lung cancer; TAK-228, dual mTORC1/2 inhibitor.
histopathological markers, with more-coherent survival profiles of the samples and/or patients
composing them. While some of the molecular clusters strongly overlap with known histologi-
cally based clusters, others surprisingly comprise samples with distinct histopathological
markers, but with similar transcriptomic profiles and survival rates. More recent analyses have
clustered multiple platforms across multiple cancer types and, as outlined above, identified
molecular similarities between tumors of different tissue of origin. For example, one study
analyzed TCGA data from >3000 samples across 12 cancer types, and found that, while most
cancers could be classified based on their histology, approximately 10% could be classified as
belonging to an ‘integrated’ subtype, that is, including cancers from multiple tissues of origin
in the same subtype [37]. Furthermore, grouping samples from different tissue types yielded
improved predictive power for patient prognosis, potentially reflecting the value of molecular
features (such as common mutations) for predicting survival [37]. The US Food and Drug
Administration (FDA) recently approved the drug pembrolizumab (an immune-checkpoint PD-1
blockade), used across many cancer types, with demonstrated effectiveness in CRC, endo-
metrial, pancreas, thyroid, and eight other cancer types, based on the presence of a specific
(mismatch-repair deficiency) signature [87]. These studies demonstrate the promise of classi-
fying tumors using molecular features, which can give additional insights into prognosis and
treatment beyond tissue of origin.

Approaches to Identifying Genetic Drivers
While as few as three to eight somatic mutations are required to drive cancer [36], identifying the
entire set of driver mutations in any given tumor is a difficult biological and computational
problem. The observation that relatively few mutations occur in a significantly recurrent manner
across tumors holds, despite the development of sophisticated statistical tools for evaluating
the significance of mutations. Researchers have developed multiple different classes of tool that
consider different information about somatic mutations, including the predicted functional
impact [88] or conservation across populations [89]. Other methods attempt to classify driver
mutations by identifying hotspots in the protein sequence or structure [90–92], or targets of
recurrent copy number aberrations [93]. Some methods also consider side information, such
as the replication timing and expression of a gene [94,95], or per patient, and/or per gene
mutation rates [96] (see references summarized in Table S2B in the supplemental information
online).
Despite these advances, in most cancer data sets, there is a ‘long tail’ of genes with infrequent
mutations, where the drivers are statistically indistinguishable from passenger mutations [32].
One report illustrated the depth of this problem by estimating the number of samples required
to detect driver mutations with a given frequency in a given cancer type through saturation
analysis [34]. The cancer type in question was critical because of the high variance in
background mutation rates in different cancers (e.g., breast cancer, prostate cancer, etc.).
For example, the authors showed that up to 5300 samples would need to be sequenced to
detect drivers occurring at 2% above the high background mutation rate in melanoma [34]. This
presents a particular challenge for rare cancer types, especially since cancer types continue to
be divided into different subtypes [34].
The observed intertumor mutational heterogeneity is widely believed to be due, in part, to

mutations targeting pathways or ‘cancer hallmarks’ [26], where each pathway includes
multiple genes, such that many different combinations of aberrations can affect hallmark
pathways and drive cancer. Thus, by uncovering the genes in these pathways, it may be
possible to identify ‘hidden’ driver mutations in the ‘long tail’, that is, the set of mutations that
are indistinguishable from passengers without considering prior knowledge, such as pathways.
To date, researchers have developed multiple classes of method that use different side
information to identify the pathways and/or hallmarks targeted in cancer (Table S2B in the
supplemental information online). One group of methods searches for significantly mutated
groups of genes in known pathway databases [97] and protein interaction networks [35,93,98–
100]. Other methods search for functional mutations that co-occur with sample-level
events [101,102], which can be viewed as a supervised learning task. These approaches
have been used on cancer cell line drug sensitivity and gene dependency and/or addiction (i.e.,
conditional essentiality) data to generate testable hypotheses, but are less well suited for
predicting coarse measurements with many factors, such as overall survival. Another promising
approach has been to search for groups of genes with mutually exclusive mutations [103–107].
However, these approaches also require large sample sizes through saturation analysis [108]
and, depending on the relative rate of driver and/or passenger mutations, such sample sizes
can be even larger than those required by the recurrent mutation detection methods
described above. Finally, going beyond coding region mutations, researchers are beginning
to uncover recurrent and functional mutations in noncoding regions of the genome that might
have a role in dysregulated gene expression, as is the case of the TERT promoter, shown for the
first time to be mutated significantly in melanoma [109] and, more recently, in 43 tumor types
[79], with significant association with poor survival in cutaneous melanoma, bladder urothelial
carcinoma, and papillary thyroid cancer [79]. Larger whole-genome sequencing efforts, such
as those from the ICGC, are likely to uncover more of these noncoding mutations due to
increased statistical power.

Integrating Genomic and Clinical Data
Current efforts linking genomic mutation data with clinical data to assist in therapeutic decisions
build and use knowledge banks (Table S3 in the supplemental information online). These
include web tools that provide data and text summaries of the frequency, mechanisms, and
druggable targets of known driver mutations [110]. Multiple tools now include ‘interpretations’
or summaries of the driver mutations written by clinicians, including the Precision Medicine
Knowledgebase (at Weill Cornell) and the Personalized Cancer Therapy knowledge base (at M.
D. Anderson), or by the ‘crowd’ [111,112] (see references in Table S2C in the supplemental
information online). A related approach recently explored leveraging existing ‘omics data sets
for the interpretation of variants in newly sequenced samples of acute myeloid leukemia [113].
For example, one study recently demonstrated the use of this approach by building survival
models that linked genomic and clinical data, and then using these models to choose treatment
(s) and predict survival for patients with newly diagnosed acute myeloid leukemia [113].
Regularized regression on both genomic and clinical features was performed on these models
to predict overall survival; the authors used these to identify additional interventions that could
increase overall survival, and extraneous interventions that could be removed for some patients
without decreasing overall survival [113]. However, effectively integrating annotations and
clinical knowledge of known variants with ‘omic databases in an automated manner for the
interpretation of patient molecular data, and creating features from molecular data for input into
survival models, remain key challenges. This appears to be largely due to the fact that most
clinical data still need to be extracted from free text, and the pertaining electronic medical
record (EMR) systems are mostly not standardized.
Identifying Cancer Vulnerabilities on the Basis of Genetic Interactions

Another way to guide precision therapy is based on identifying and utilizing genetic inter-
actions, in particular, by harnessing the concept of synthetic lethal interactions (SLi). SLi
describe the relationship between two genes where an individual inactivation of either gene
results in a viable phenotype, while their combined inactivation is lethal for the cancer cell
[114,115]. SLi have long been considered a foundation for the development of selective
anticancer therapies [64,114,115], which aim to inhibit the SL partner of a gene that is
inactivated de novo in cancer cells. Given that this SLi partner gene is most likely to be inhibited
only in the tumor, this treatment will primarily kill these cancer cells but not healthy ones. Thus,
this offers a complementary approach for predicting patient drug responses to sequence-
based cancer precision medicine strategies. This might be achieved by: (i) going beyond
existing precision oncology approaches based on actionable mutations (i.e., a mutation that
can be targeted by specific small-molecule inhibitors) in a few hundred cancer driver genes, and
examining the whole genome, thereby covering all possible changes that might have occurred
in a tumor. These might uncover more treatment options for patients whose tumors do not bear
actionable mutations; and (ii) SLi are well poised to offer effective options for potentially treating
heterogeneous tumors, impacting differentiating subclones and resulting overall in more
effective tumor eradication with a reduced likelihood of drug resistance.
Given this promising potential, extensive experimental efforts have aimed to tease out the wiring
of genetic interactions in cancer cells based on single (isogenic) cell lines [116–121] or on large-
scale genetic knockout-based screens [122–126]. However, due to the large combinatorial
space of pairwise interactions that need to be surveyed, these screens have probed only a
small fraction of the coverage offered by SLi: each screen typically scans a few thousand
candidate SL partners of just one ‘anchor’ cancer driver gene of interest (e.g., KRAS or VHL),
altogether covering a mere fraction of the 500 million gene pairings in the human genome. Yet,
with these screens, several SL interactions have been successfully uncovered to date; apart
from examining the effect of PARP inhibitors in patients with BRCA-mutated breast and

pancreatic tumors, a growing number of other treatments targeting SL-based cancer specific
vulnerabilities are currently being clinically investigated [127].
Aiming to bypass the limitations of current experimental techniques in probing the vast space of
potential SLi, various computational approaches have been developed to identify such candi-
date SLi (see references in Table S2D in the supplemental information online). These include
applying various machine-learning methodologies to predict genetic interactions in different
species [128–131], and in cancer (using yeast SLi) [119,132], utilizing metabolic modeling
[133,134], evolutionary characteristics [119,129], and transcriptomic profiles [101,135], and,
more recently, by mining cancer patient data [136–138] (Table S2D in the supplemental
information online). One recent study evaluated the TCGA copy number and transcriptomics
data to identify as candidate SLis, gene pairs that are almost never found inactivated in the
same tumors [136]. The study demonstrated that gene pair interactions (a subset of which was
validated in experimental screens) could be successfully used to predict the survival of patients
with breast cancer in an independent data set [136]. The pair was also used to predict in vitro
drug responses to identify novel drug repurposing indications for potentially treating renal
cancer [136]. Unlike the approach of using expression and copy number data [136], an
algorithm was recently developed to mine pan-cancer human tumor data and define muta-
tion-specific SL interactions for specific cancers [139]. Its SL predictions were validated against
published SL screens and one specific SL gene pair interaction between mutated IDH1 and the
gene encoding acetyl-CoA carboxylase 1 (ACACA) in leukemia was experimentally validated;
this interaction attenuating tumor growth in patient-derived xenografts(PDX) [139]. Finally,
certain predicted SL interactions were shown to successfully predict drug sensitivity, thus
serving as biologically interpretable biomarkers of the latter [139]. Overall, while these studies
have laid a solid basis for some of these genome-wide approaches, extensive research is
warranted to further elucidate the potential of SLi-based approaches in precision oncology.
Moreover, for clinical trials, there is an important unmet need to specifically design and test SLi-
based approaches that may uncover a range of tumor-specific vulnerabilities.
The Use of In Vitro and In Vivo Models for Guiding Precision Therapy
The increased functional annotation of genetic variants of unknown significance [140], along
with systematic high-throughput drug screens in 1000 cancer cell lines [141] or 1000 PDX
models [142], has significantly increased our understanding of the relationship between
genotype and drug sensitivity. We are beginning to understand the molecular profiles of
patients responding to conventional chemotherapy, such as to temozolomide [143] and
other DNA-damaging agents [144]. Due to their well-characterized clinical benefit in unstratified
patient cohorts, these remain equally valuable therapy choices in addition to targeted therapies.
With this ever-increasing number of validated biomarkers and available drugs, it is expected
that molecular profiling will reveal multiple potentially actionable alterations, which may be
treated with a multitude of drugs and/or drug combinations. Prioritizing predicted treatments
requires functional testing, especially in cases where the drug–biomarker association has not
been clinically validated. Undoubtedly, there will still be patients whose molecular analysis is
either not feasible or does not reveal targetable alterations, for which alternate routes to inform
therapy are necessary. For this purpose, several in vitro or in vivo patient-derived functional
platforms (e.g., PDX or organoid models) have been developed that mimic the native features of
tumors more closely than conventional cell culture drug-screening platforms [145].
PDX models offer one attractive approach, because tumor heterogeneity is maintained in these
models at least in early passages (for a comprehensive review, see [146]). In addition, clinical
studies have demonstrated remarkable correlations between drug activity in the PDX model
and a patient’s clinical outcome [146–149]. However, not all human tumor samples grow in
mice following subcutaneous or orthotopic implantation, and the long time span needed for

tumor development and expansion to test multiple drugs and/or drug regimens restricts this
approach to patients with a less aggressive disease course [146]. Serial passaging is not only
accompanied by the substitution of human stroma with murine components, eventually
affecting clonal evolution [146], but also results in extensive mouse colonies and, hence,
logistical difficulties and rapidly expanding costs. Finally, although humanized mouse models
with a (partially) competent ‘human’ immune system have been developed, the remaining
technical and biological difficulties of generating these mice restrict the use of PDX models in
studying immunotherapeutic approaches as well as the effects of immunity on the efficacy of
other drugs in preclinical models [146]. Nonetheless, alternate, in vitro or ex vivo models may
substitute the extensive use of PDX.
Patient-derived 3D organoids provide a practical alternative (see other models discussed in Box
2). Organoids are established by dissociating and embedding tissue in an cell-free extracellular
matrix (matrigel or collagen), which can be expanded in a growth factor-enriched medium [150].
Organoids from pancreatic [151], colon [152–154], gastric [155], prostate cancer [156], and
brain tumors and/or metastasis [157] have been established, and have the advantage of 3D
growth of normal and cancer tissue, recapitulating copy number and mutation spectra, as well
as other physiologically relevant aspects of disease progression in vitro [150–158]. Organoids
can be established in culture from needle biopsies within a relative short time period, and have
also been generated from circulating tumor cells(CTCs) [156]. Organoids can serve as a
model system to perform high-throughput screens within a clinically relevant time frame: in a
larger precision oncology study, organoids were established from fresh tissue available from
38% of 145 patients [158]. In addition, PDX models were successfully established from these
organoids in 19 of 22 attempts [158]. High-throughput drug screens were performed (160
drugs, including chemotherapy and targeted therapy) in 2D cultures from four patients, and the
best ‘hits’ (drugs that most effectively decreased cell viability in vitro) were verified in 3D
organoid cultures. Selected treatments were then tested in combination to identify effective
combination therapies. The best hits of single and combination therapies from two patients
were further tested in 3D organoids and PDX models, validating tumor responses in vivo, and
compared with the efficacy of current patient treatments [158]. In addition, potential drug
Box 2. Valuable Models for Guiding Precision Therapy

In addition to PDX and organoids, conditional reprogramming(CR) of patient-derived primary epithelial tumor cells or
organotypic cultures are among the possibilities to test selected treatments. Patient-derived cell lines via CR can be
rapidly established [210] and are suitable to screen large drug libraries [211,212], or to test drug combinations to
overcome acquired resistance to targeted therapy [213]. While phenotypic features and the genetic heterogeneity of the
original tumor are retained in short-term CR cultures, the enrichment of specific cell populations, including nontrans-
formed epithelial cells in this model, requires the crossverification of pheno- and genotypic features of donor tissues and
CR cells. The lack of a 3D environment may be partially overcome by culturing CR cells in sophisticated 3D artificial
organotypic cultures [214], of which fully automated 1536-well high-throughput screening platforms have recently been
described [215]. Although these artificial microenvironments lack the heterogeneity observed in patient tumors, they
may allow testing tumor cell behaviors in the context of different organ microenvironments shown to influence drug
responses [215].
In organotypic slice cultures [216] or organ explants [217], either thin slices of the tumor sample or minced tumor tissues
are maintained in culture. The biggest advantage of these culture types is that they retain cancer associated stromal
cells, preserve tumor–stroma interactions, signaling pathways, and gene expression profiles [218]. Improvements
include the use of autologous serum and patient-specific stromal-matrix proteins to more closely resemble individual
microenvironmental conditions [219], aiming to accurately predict responses to anticancer drugs. However, not all
tissues are suitable to generating thin slices (e.g., soft, mucinous, or fatty tissue), where firm tissue consistency is
required [220,221]. Another drawback of slice cultures is the loss of viability within 5–7 days [218]. Given that the median
time frame for molecular profiling and data processing in precision oncology trials is 2–4 weeks, the method is not
suitable for testing genomics-guided therapies derived from the same biopsy, unless combined with other models.
Recently, organotypic slice cultures established from pancreatic ductal adenocarcinoma PDX models were used to
screen against clinically relevant drug regimens in a 96-well format, demonstrating consistency between sensitivity of
organotypic cultures and the clinical responses of donor patients [222].

toxicities were evaluated (e.g., trametinib and afatinib led to significant weight loss in mice)
[158]. For both cases, the combination of targeted therapies was superior over standard
chemotherapy [158]. This study underscores the potential use of functional screens in patients
where no targeted therapies are available, and the possibility of identifying effective drugs and/
or drug combinations [158]. The study further demonstrated that therapy recommendations
could be retrieved within a clinically relevant time frame (between 7 and 13 weeks) [158],
highlighting the importance of defining regulatory routes that might simplify off-label drug
access for late-stage patients, who are often not eligible for clinical trial enrollment (see
Outstanding Questions). Therefore, the combination of molecular profiling (genomics and
transcriptomics) and functional testing holds promise for determining effective combination
therapies for individual patients with cancer.
Clinical Management of Therapy Resistance in the Precision Oncology Era

Although genomics-guided therapy is associated with prolonged progression-free survival,
patients with cancer can generally develop resistance to drugs within 6–12 months, even when
trunk mutations are targeted (Figure 2). Such acquired resistance to targeted drugs may be
explained by the selection of resistant cancer cells that are present before therapy or that are
generated de novo as a result of genomic instability. Sequential therapy of second-, third-, and
even fourth-generation inhibitors that specifically address emerging mutations within the
original target (i.e., EGFR inhibitors [159–163]) or drugs targeting newly established driver
mutations (e.g., MET amplification in EGFR inhibitor-resistant CRCs [164]) have been used to
overcome resistance. Alternatively, combined inhibition of multiple pathways [165] or vertical
pathway inhibition (targeting multiple proteins within one pathway) have been suggested to
prolong progression-free survival by pre-empting resistance in a pro-active manner and
inhibiting the selection of resistant clones [166]. As an example, targeted therapy has been
used against BRAF and MEK in melanoma to counteract feedback regulatory loops and
achieve efficient pathway inhibition [18]. Although these approaches are suitable to prolong
progression-free survival, management of resistant disease remains dismal and/or short-lived,
partly because multiple resistance mutations (in addition to other mechanisms) can occur
simultaneously. This heterogeneity was recently demonstrated in a patient with CRC, where a
MEK1 mutation was detected in a liver metastasis biopsy, conferring resistance to cetuximab
[167]. Treatment of this patient with the combination of panitumumab and trametinib resulted
in regression of the biopsied liver metastasis; however, other liver metastases progressed while
on treatment. Analysis of circulating tumor DNA(ctDNA) revealed a previously unrecognized
KRAS mutation, which was later found in a biopsy from a progressing liver metastasis,
highlighting the challenges of combating polyclonal resistance [167]. To address these issues,
more general approaches have been suggested, such as interfering with tumor evolutionary
programs, for instance, by increasing genomic instability to lethal levels (e.g., PARP inhibitors in
tumors with homologous recombination deficit) [25]. These have been promising strategies to
exploit genome instability, as one driving force of heterogeneity, therapeutically [25]. However,
even this approach is accompanied by resistance [25]. Finally, there is growing interest in
applying intermittent treatment doses, or adjusting drug doses to limit the evolutionary
pressure imposed by a given drug [25]. Such adaptive therapy may serve to maintain a
drug-sensitive population, with the goal of stabilizing the tumor size rather than eliminating the
tumor. Preliminary evidence for the putative benefit of such drug ‘holidays’ comes from patients
with CRC receiving EGFR therapy [168], which was also suggested for melanoma [169,170]
and, recently, breast cancer models [171]. Advances in the characterization of ctDNA now
make it possible to carefully evaluate these different methods to combat genetic resistance to
targeted drugs in the clinical setting [167,168,172–175]. Furthermore, the reappearance of the
driver mutation or the appearance of previously undetected mutations associated with resis-
tance to targeted therapy in the blood can enable early detection of therapy failure (before tumor

Figure 2. Targeted Therapy and Mechanisms of Acquired Resistance. Major classes of current US Food and
Drug Administration (FDA)-approved targeted therapies include (A) drugs targeting oncogenic drivers or drugs targeting
other genetic vulnerabilities, such as poly (ADP-ribose) polymerase (PARP) inhibitors in tumors with homologous
recombination (HR) deficiency; (B) drugs that aim to increase the antitumor immune response or (C) inhibit neoangiogen-
esis. Numerous genetic as well as nongenetic mechanisms (green boxes) of acquired resistance to targeted therapeutics
are known, which likely act in concert to mediate the largely short-lived response to these drugs. (D) Routes to monitor
emerging resistance as well as the suitability of liquid biopsies compared with tumor biopsies to inform second-line therapy
are displayed. Abbreviations: APC, antigen-presenting cell; CAFs, cancer-associated fibroblasts; CTC, circulating tumor
cell; ctDNA, circulating tumor DNA; EMT, epithelial–mesenchymal transition; NSCLC, non-small cell lung cancer; RTK,
receptor tyrosine kinase; SCLC, small cell lung cancer.

imaging indicates relapse [173,174]), and might also identify new potential drivers suitable for
guiding second-line therapy [167,168,172–175], a promising approach for the management of
resistant disease.
In addition to the Darwinian-like evolution of genomic alterations in resistant clones under

therapeutic pressure, tumors evolve resistance to therapy by adaptively rewiring transduction
networks to support the signaling processes required for survival and/or tumor maintenance
in a postgenomic, transient, and dynamic manner [176] (Figure 2; Tables S2F and S3 in the
supplemental information online). One such example is the ability of BRAF-inhibitor sensitive
MITFhigh/AXLlow melanoma cell populations to readily switch into a MITFlow/AXLhigh drug-
resistant population [177,178]; these cells have been shown by single-cell RNA-seq to pre-
exist in treatment-naïve samples [179]. Another example of such phenotypic plasticity was
reported in ER+HER2 breast cancers [180]. Following multiple courses of therapy, HER2+
cells lacking gene amplification have been found to emerge in addition to HER2 cells in
patient tumors and among CTCs; this may be indicative of nongenetic mechanisms involved
in HER2 upregulation [180]. Furthermore, characterization of patient-derived CTCs revealed
that, although both HER2+ and HER2 CTCs maintained tumor-initiating potential in ortho-
topic xenograft experiments, HER2+ cells were highly proliferative and sensitive to chemo-
therapy, whereas HER2 CTCs exhibited a slow proliferation rate, upregulated NOTCH
signaling, and were chemoresistant [180]. Of note, cells could interconvert between a
HER2+-chemosensitive, and a HER2 -chemoresistant state, and, with chemotherapy, a
HER2+ population could shift towards a HER2 phenotype [180]. Accordingly, targeting
NOTCH in combination with chemotherapy (paclitaxel) suppressed tumor growth in mice,
whereas either treatment alone was inefficient in limiting tumor growth [180]. Therefore, rapid
interconversion of CTCs between distinct functional states may contribute to acquired
resistance to therapy; consequently, it is possible that combination therapy might improve
therapeutic responses and delay the onset of resistance. Other nongenetic routes to escape
targeted therapy can include the epithelial–mesenchymal transition (EMT), a developmental
program that is often hijacked by cancer cells [181]. Preclinical models have associated the
EMT program with chemoresistance [182,183] and a subset of patients with non-small cell
lung cancer (NSCLC) resistant to EGFR-targeted therapy were shown to display an increase
in ‘mesenchymal’ cancer cells [184]. Additionally, tumors can also undergo ‘histological
transformations’, as shown for patients with EGFR inhibitor-resistant NSCLC whose tumors
converted to a small cell lung cancer (SCLC) phenotype, escaping therapy [184]. Finally,
vascular mimicry, a phenomenon where tumor cells transdifferentiate into endothelial-like
cells that can form matrix-rich, vascular-like, perfused channels, has also been proposed to
contribute to resistance to antiangiogenic therapy, but further testing of this mechanism is
required to better understand its role in resistance [185].
Multiple resistance mechanisms (genetic and nongenetic) can act in concert to confer resis-
tance to targeted therapy (Figure 2). Indeed, in a recent melanoma study [186], analysis of
patient-matched melanoma tumors biopsied before therapy and during disease progression
demonstrated that, in contrast to heterogeneous genetic mechanisms that could result in
acquired resistance, transcriptomic signatures were highly recurrent in serial biopsies; these
indicated that a multitude of genetic and epigenetic events within the tumor compartment
converged on specific genes (c-MET, LEF1, and YAP1) and pathways to mediate resistance,
consistent with a canalization evolutionary process [186]. Additionally, this acquired
resistance signature correlated with changes in the tumor immune microenvironment, including
depletion of intratumoral CD8+ T cells, exhaustion of tumor-reactive CD8+ T cells, and loss of
antigen presentation; these have been previously linked to resistance to anti-PD-1 salvage
therapy in biopsies of patients with melanoma, in support of the presumed role of CD8+ T cell
exhaustion in the development of resistance [77,186]. Furthermore, these findings suggest that

first-line therapy with immune-checkpoint inhibitors followed by BRAF-targeted therapy upon
relapse in BRAF-mutant melanomas, may be superior over BRAF-inhibitor frontline therapy, a
hypothesis that is currently being tested in a clinical trial (NCT02224781).
The mere follow-up of genomic alterations in ctDNA is unlikely to result in the satisfactory
management of resistant disease because adaptive mechanisms on transcriptional and
signaling levels can be missed by genomic analysis alone (Figure 2). The use of CTCs may
be preferred to monitor resistant disease, because they have a better overall prognostic value,
provide an additional opportunity to characterizing genetic and nongenetic ITH, and are
suitable for functional studies [187]. The previously described in vitro and/or ex vivo models
may serve to detect relevant signaling nodes and counteract adaptive signaling by combina-
torial therapy to delay onset of resistance. As suggested for the selection of first-line therapies, a
comprehensive genomic, transcriptomic, and functional analysis of resistant disease is
required to overcome these challenges. In that case, it may be possible to mine the increasingly
available ‘omics data from large cohorts of patients and identify genetic interactions that
mediate network-wide signaling alterations conferring resistance. Such an approach could
exploit the much less-studied type of genetic interactions, termed synthetic rescues (SRs),
also known as suppression interactions [188–192]. Such SRs denote a functional interac-
tion between two genes where the targeting of one gene is compensated by the altered activity
of another gene (termed the ‘rescuer gene’); this can restore and rescue cell fitness, leading to
drug resistance. Similar to SLs, SR interactions could in principle be identified by mining ‘omics
data from large cohorts of pretreated tumor samples, taking into consideration that functional
alterations might already be occurring during the natural evolution of tumorigenic populations.
Concluding Remarks: The Road Ahead

Despite the limitations of genomics-driven cancer therapy, current efforts have demonstrated
that this approach has the potential to improve clinical outcomes, although at this time, only for
a minority of patients. Future precision oncology treatment will need to include the broader
landscape of genetic and epigenetic changes that take place in a tumor. Those include the
tumor microenvironment, which comprises metabolic as well as immunological changes, in
addition to the influence exerted by the microbiome. Consequently, we must understand that
targeting a single pathway in a tumor is, in most cases, not sufficient to achieve a sustained
response. We have outlined how the inclusion of transcriptomic data could serve to stratify
patients, suggest combination therapies, or define novel vulnerabilities to improve upon current
precision oncology trials (Figure 3 and Box 3). Patient-derived ex vivo and/or in vivo models can
serve to identify the toxicity and efficacy of combination therapies, link genomics to drug
responses, and, when such information is included in a mineable database, could serve to
inform therapeutic decision-making. The limiting factor in performing multiple ‘omics
approaches and functional testing, aside from cost, is tissue availability, highlighting the need
to improve methods to retrieve sufficient tumor material. As such, the use of CTCs is appealing,
because these can be non-invasively isolated and may better reflect the prevailing ITH (see
Outstanding Questions).
At present, ITH appears to be the biggest obstacle we need to overcome to achieve a sustained
therapeutic response. With rapid technological advances, we are acquiring the toolbox to
comprehensively characterize the complex heterogeneity of tumors at the single cell level (Box
4). However, computing the data from different platforms of 1000 cells or more remains a
challenge. Among the questions that emerge is whether it will be possible to develop multiplex-
based approaches, or use machine-learning techniques to compute tumor trends to predict
subcluster and clonal behaviors (see Outstanding Questions).

Figure 3. Precision Oncology Work-
flow. In this proposed precision oncology
workflow, patient molecular profiles are
used to suggest first-line (combination)
therapies. Functional models serve to test
the safety and efficacy of selected drugs
or screen for drugs in cases when mole-
cular profiling is not informative. Patients
are closely monitored during the course
of therapy to detect resistance. Molecular
profiling, computational analysis, and
functional testing of resistant tumors are
repeated upon onset of resistance to
determine second-line therapy options.
Abbreviations: PDX, patient-derived
xenografts.
Among the first series of steps needed to provide important insights into the complex nature of
malignancies, the inclusion of technological advances and computational approaches in clinical
trials stands out. We should aim to map and analyze the impact of complex genetic and
nongenetic heterogeneity during cancer progression and therapy regimens to help pave the
Box 3. Clinicians Corner

Targeted drugs are largely based on defined drugs (small molecules or biologics) designed to inhibit specific oncogenic
mutations or target key regulatory nodes that drive tumorigenesis or underlie cancer vulnerabilities. Usually, the
presence of drug-specific biomarkers enables stratification of patients for therapy and monitoring drug effectiveness.
Given the success of targeted therapies, together with the recognition that different tumor types share driver and/or
master regulators, the use of drugs that target common regulatory nodes in a histology-agnostic manner is being
evaluated in clinical trials.
Clinical experience with genomics-guided cancer therapy supports the notion that genomic profiling can improve
patient outcomes. The degree of success can be associated with the ability to verify the role of a targeted mutation and/
or alteration in tumor development, or vice versa; that is, whether the drug can efficiently attenuate the tumorigenic
effects orchestrated by the genetic alteration.
Precision oncology cannot be limited to genetics to predict responses to therapy, neither can it be limited to a single
‘omics-based approach. Multiple drivers can underlie tumor heterogeneity, which in turn can confer resistance,
metastasis, and dormancy. It also requires the targeting of master regulators that are influenced by epigenetic and
microenvironmental-based pathways. The inclusion of additional platforms, of which at this time the most suitable
appears to be transcriptomics, (with future inclusion of metabolome and microbiome analysis), is highly desirable to
identify such master regulators and design more-precise therapeutic modalities.
Advances in computational approaches to mine and integrate the multitude of data sets that become available to us are
expected to allow better subclassification of patient cohorts into subpopulations able to respond to a given therapy. In
addition, integration of multiple data platforms is expected to drive the identification of novel vulnerabilities, which will
further add to the armamentarium of current anticancer therapies. This may likely result in newer stratification methods
to identify patients who might benefit from a given precision oncology approach.
The implementation of powerful ex vivo or improved in vivo PDX models in the planning of clinical practice is encouraged.
By using these models, a multitude of available drugs and predictive biomarkers might be assessed to evaluate
therapeutic options, as well as their combinations and delivery sequence and/or approaches.

Box 4. Available Tools to Decipher Intratumor Heterogeneity
Outstanding Questions
ITH manifests as differences in genetic, epigenetic, and signaling networks of individual tumor cells coupled with
Can the integration of transcriptomic
heterogeneity within the stromal compartment [25,27]. While ITH is influenced by the inherent tumor genetic make-up,
information in clinical decision-making
epigenetic states (influenced by the location of tumor cells) and microenvironmental factors have been recognized as
improve patient outcomes? There is a
being equally important in dictating the diverse cellular states that drive the primary tumor or its metastatic lesions.
strong biological rationale to support
Those include the proximity to endothelial cells, cancer-associated fibroblasts, immune cells, as well as the nutrient and
this hypothesis. However, the neces-
oxygen availability and biophysical properties of the extracellular matrix [28,207]. The development of single cell
sary steps to enable integration of dis-
separation and analysis methods has provided critical insights into the complex heterogeneity of tumors, where
tinct data platforms and to rigorously
multiple clusters of genetically [223] and (more so) phenotypically distinct populations exist. The resulting cell-to-cell
test the clinical value of transcriptom-
variability in stemness and differentiation programs, proliferation. and quiescence markers, as well as in the expression
ics remain to be assessed.
of predictive biomarkers [179,180,224,225] define the propensity of a tumor for therapeutic resistance, metastasis, and
dormancy.
How many platforms within the tumor
microenvironment (immune, micro-
Active research is being undertaken to overcome limitations in single cell methods. These include (i) improving
biome, and metabolome) need to be
throughput of single cell DNA- [226–228] and single cell RNA-seq [229–231] methods; (ii) limiting sampling bias,
integrated to provide a multidimen-
which can partially be resolved with liquid biopsies [187]; and (iii) minimizing tissue dissociation and disruption of spatial
sional map of the complex tumor land-
information [232].
scape? Will it allow a more accurate
prediction of regulatory nodes and
possible therapeutic modalities?
way towards the next generation of cancer therapeutics, capable of overcoming the challenges
imposed by the defying intratumor heterogeneity of cancers. Preclinical patient-derived models
(PDX and/or organoids) demonstrate
a strong correlation with patient out-
Acknowledgments comes. Can regulatory guidelines be
Support by NCI grant (R35CA197465-01), and a Hervey Family Non-endowment Fund at the San Diego Foundation (to Z. defined to render the use of off-label
R.), are gratefully acknowledged. drugs in the clinic easier when based
on positive outcomes taken from such
Supplemental Information models?
Supplemental information associated with this article can be found online at http://dx.doi.org/10.1016/j.molmed.2017.08.
How many single cells (and from how
003.
many tumors) will we need to
sequence to accurately depict the
References complex heterogeneity observed
1. Vargas, A.J. and Harris, C.C. (2016) Biomarker development in 14. Saltz, L.B. et al. (2004) Phase II trial of cetuximab in patients with
within patients? Can machine-learning
the precision medicine era: lung cancer as a case study. Nat. refractory colorectal cancer that expresses the epidermal
Rev. Cancer 16, 525–537 growth factor receptor. J. Clin. Oncol. 22, 1201–1208
approaches be used to predict the
behavior of a tumor?
2. Hyman, D.M. et al. (2017) Implementing genome-driven oncol- 15. Ledermann, J. et al. (2014) Olaparib maintenance therapy in
ogy. Cell 168, 584–599 patients with platinum-sensitive relapsed serous ovarian cancer:
3. Prasad, V. (2016) Perspective: the precision-oncology illusion. a preplanned retrospective analysis of outcomes by BRCA Are CTCs a reliable source to compre-
Nature 537, S63 status in a randomised phase 2 trial. Lancet Oncol. 15, 852–861
hensively map tumor heterogeneity
4. Tannock, I.F. and Hickman, J.A. (2016) Limits to personalized 16. Cancer Genome Atlas, N (2015) Genomic classification of cuta- and do they accurately define pheno-
cancer medicine. N. Engl. J. Med. 375, 1289–1294 neous melanoma. Cell 161, 1681–1696
typic heterogeneity of advanced dis-
5. Voest, E.E. and Bernards, R. (2016) DNA-guided precision 17. Hauschild, A. et al. (2012) Dabrafenib in BRAF-mutated meta-
ease? How can we improve the
medicine for cancer: a case of irrational exuberance? Cancer static melanoma: a multicentre, open-label, phase 3 random-
ised controlled trial. Lancet 380, 358–365 isolation of these valuable cells?
Discov. 6, 130–132
6. Karnoub, A.E. and Weinberg, R.A. (2008) Ras oncogenes: split 18. Johnson, D.B. et al. (2016) Emerging targeted therapies for
personalities. Nat. Rev. Mol. Cell Biol. 9, 517–531 melanoma. Expert Opin. Emerg. Drugs 21, 195–207 How can ITH be exploited
7. Sidransky, D. (1997) Nucleic acid-based methods for the detec- 19. Tang, X.H. and Gudas, L.J. (2011) Retinoids, retinoic acid therapeutically?
tion of cancer. Science 278, 1054–1059 receptors, and cancer. Annu. Rev. Pathol. 6, 345–364
8. Soussi, T. (2010) The history of p53: a perfect example of the 20. Jones, P.A. et al. (2016) Targeting the cancer epigenome for
drawbacks of scientific paradigms. EMBO Rep. 11, 822–826 therapy. Nat. Rev. Genet. 17, 630–641
9. Chapman, P.B. et al. (2011) Improved survival with vemurafenib 21. Ferrara, N. and Kerbel, R.S. (2005) Angiogenesis as a thera-
in melanoma with BRAF V600E mutation. N. Engl. J. Med. 364, peutic target. Nature 438, 967–974
2507–2516 22. Miller, J.F. and Sadelain, M. (2015) The journey from discoveries
10. Cobleigh, M.A. et al. (1999) Multinational study of the efficacy in fundamental immunology to cancer immunotherapy. Cancer
and safety of humanized anti-HER2 monoclonal antibody in Cell 27, 439–449
women who have HER2-overexpressing metastatic breast can- 23. Cancer Genome Atlas Research, N et al. (2013) The Cancer
cer that has progressed after chemotherapy for metastatic Genome Atlas Pan-Cancer analysis project. Nat. Genet. 45,
disease. J. Clin. Oncol. 17, 2639–2648 1113–1120
11. Druker, B.J. et al. (2001) Efficacy and safety of a specific inhibitor 24. International Cancer Genome, C et al. (2010) International net-
of the BCR-ABL tyrosine kinase in chronic myeloid leukemia. N. work of cancer genome projects. Nature 464, 993–998
Engl. J. Med. 344, 1031–1037 25. McGranahan, N. and Swanton, C. (2017) Clonal heterogeneity and
12. Kwak, E.L. et al. (2010) Anaplastic lymphoma kinase inhibition in tumor evolution: past, present, and the future. Cell 168, 613–628
non-small-cell lung cancer. N. Engl. J. Med. 363, 1693–1703 26. Hanahan, D. and Weinberg, R.A. (2011) Hallmarks of cancer:
13. Lynch, T.J. et al. (2004) Activating mutations in the epidermal the next generation. Cell 144, 646–674
growth factor receptor underlying responsiveness of non-small- 27. Marusyk, A. et al. (2012) Intra-tumour heterogeneity: a looking
cell lung cancer to gefitinib. N. Engl. J. Med. 350, 2129–2139 glass for cancer? Nat. Rev. Cancer 12, 323–334

28. Tabassum, D.P. and Polyak, K. (2015) Tumorigenesis: it takes a biomarker and early clinical evaluation in Western and Japanese
village. Nat. Rev. Cancer 15, 473–483 patients with advanced solid tumors. Cancer Res. 73 (8 Suppl),
29. Hofree, M. et al. (2016) Challenges in identifying cancer genes by LB–66
analysis of exome sequencing data. Nat. Commun. 7, 12096 54. Cappuzzo, F. et al. (2006) HER2 mutation and response to
30. Forbes, S.A. et al. (2017) COSMIC: somatic cancer genetics at trastuzumab therapy in non-small-cell lung cancer. N. Engl. J.
high-resolution. Nucleic Acids Res. 45, D777–D783 Med. 354, 2619–2621
31. Futreal, P.A. et al. (2004) A census of human cancer genes. Nat. 55. Hammerman, P.S. et al. (2011) Mutations in the DDR2 kinase
Rev. Cancer 4, 177–183 gene identify a novel therapeutic target in squamous cell lung
cancer. Cancer Discov. 1, 78–89
32. Garraway, L.A. and Lander, E.S. (2013) Lessons from the can-
cer genome. Cell 153, 17–37 56. Iyer, G. et al. (2012) Genome sequencing identifies a basis for
everolimus sensitivity. Science 338, 221
33. Hofree, M. et al. (2013) Network-based stratification of tumor
mutations. Nat. Methods 10, 1108–1115 57. Liao, R.G. et al. (2013) Inhibitor-sensitive FGFR2 and FGFR3
mutations in lung squamous cell carcinoma. Cancer Res. 73,
34. Lawrence, M.S. et al. (2014) Discovery and saturation analysis of
5195–5205
cancer genes across 21 tumour types. Nature 505, 495–501
58. Serra, V. et al. (2013) Clinical response to a lapatinib-based
35. Leiserson, M.D. et al. (2015) Pan-cancer network analysis iden-
therapy for a Li-Fraumeni syndrome patient with a novel
tifies combinations of rare somatic mutations across pathways
HER2V659E mutation. Cancer Discov. 3, 1238–1244
and protein complexes. Nat. Genet. 47, 106–114
59. Subbiah, V. et al. (2014) Targeted therapy by combined inhibi-
36. Vogelstein, B. et al. (2013) Cancer genome landscapes. Science
tion of the RAF and mTOR kinases in malignant spindle cell
339, 1546–1558
neoplasm harboring the KIAA1549-BRAF fusion protein. J.
37. Hoadley, K.A. et al. (2014) Multiplatform analysis of 12 cancer Hematol. Oncol. 7, 8
types reveals molecular classification within and across tissues
60. Van Allen, E.M. et al. (2015) Genomic correlate of exceptional
of origin. Cell 158, 929–944
erlotinib response in head and neck squamous cell carcinoma.
38. Hartmaier, R.J. et al. (2017) Genomic analysis of 63,220 tumors JAMA Oncol. 1, 238–244
reveals insights into tumor uniqueness and targeted cancer
61. Wagle, N. et al. (2014) Response and acquired resistance to
immunotherapy strategies. Genome Med. 9, 16
everolimus in anaplastic thyroid cancer. N. Engl. J. Med. 371,
39. Dienstmann, R. et al. (2015) Database of genomic biomarkers 1426–1433
for cancer drugs and clinical targetability in solid tumors. Cancer
62. Wagle, N. et al. (2014) Activating mTOR mutations in a patient
Discov. 5, 118–123
with an extraordinary response on a phase I trial of everolimus
40. Rubio-Perez, C. et al. (2015) In silico prescription of anticancer and pazopanib. Cancer Discov. 4, 546–553
drugs to cohorts of 28 tumor types reveals targeting opportu-
63. Chen, J.C. et al. (2014) Identification of causal genetic drivers of
nities. Cancer Cell 27, 382–396
human disease through systems-level analysis of regulatory
41. Biankin, A.V. et al. (2015) Patient-centric trials for therapeutic networks. Cell 159, 402–414
development in precision oncology. Nature 526, 361–370
64. Luo, J. et al. (2009) Principles of cancer therapy: oncogene and
42. Rashdan, S. and Gerber, D.E. (2016) Going into BATTLE: non-oncogene addiction. Cell 136, 823–837
umbrella and basket clinical trials to accelerate the study of
65. Alizadeh, A.A. et al. (2015) Toward understanding and exploiting
biomarker-based therapies. Ann. Transl. Med. 4, 529
tumor heterogeneity. Nat. Med. 21, 846–853
43. Le Tourneau, C. et al. (2016) Treatment algorithms based on
66. Robinson, D. et al. (2015) Integrative clinical genomics of
tumor molecular profiling: the essence of precision medicine
advanced prostate cancer. Cell 161, 1215–1228
trials. J. Natl. Cancer Inst. 108, djv362
67. Fishbein, L. et al. (2017) Comprehensive molecular characteri-
44. Massard, C. et al. (2017) High-throughput genomics and clinical
zation of pheochromocytoma and paraganglioma. Cancer Cell
outcome in hard-to-treat advanced cancers: results of the
31, 181–193
MOSCATO 01 trial. Cancer Discov. 7, 586–595
68. Kato, S. et al. (2017) RET Aberrations in diverse cancers: next-
45. Le Tourneau, C. et al. (2015) Molecularly targeted therapy based
generation sequencing of 4,871 patients. Clin. Cancer Res. 23,
on tumour molecular profiling versus conventional therapy for
1988–1997
advanced cancer (SHIVA): a multicentre, open-label, proof-of-
concept, randomised, controlled phase 2 trial. Lancet Oncol. 69. Stransky, N. et al. (2014) The landscape of kinase fusions in
16, 1324–1334 cancer. Nat. Commun. 5, 4846
46. Wheler, J.J. et al. (2016) Cancer Therapy directed by compre- 70. Alvarez, M.J. et al. (2016) Functional characterization of somatic
hensive genomic profiling: a single center study. Cancer Res. mutations in cancer using network-based inference of protein
76, 3690–36701 activity. Nat. Genet. 48, 838–947
47. Lopez-Chavez, A. et al. (2015) Molecular profiling and targeted 71. Tian, S. et al. (2013) A combined oncogenic pathway signature
therapy for advanced thoracic malignancies: a biomarker- of BRAF, KRAS and PI3KCA mutation improves colorectal
derived, multiarm, multihistology phase II basket trial. J. Clin. cancer classification and cetuximab treatment prediction. Gut
Oncol. 33, 1000–1007 62, 540–549
48. Stockley, T.L. et al. (2016) Molecular profiling of advanced solid 72. Vecchione, L. et al. (2016) A vulnerability of a subset of colon
tumors and patient outcomes with genotype-matched clinical cancers with potential clinical utility. Cell 165, 317–330
trials: the Princess Margaret IMPACT/COMPACT trial. Genome 73. Lord, C.J. and Ashworth, A. (2016) BRCAness revisited. Nat.
Med. 8, 109 Rev. Cancer 16, 110–120
49. Hyman, D.M. et al. (2015) Vemurafenib in multiple nonmelanoma 74. Konstantinopoulos, P.A. et al. (2010) Gene expression profile of
cancers with BRAF V600 mutations. N. Engl. J. Med. 373, BRCAness that correlates with responsiveness to chemother-
726–736 apy and with outcome in patients with epithelial ovarian cancer.
50. Kopetz, S. et al. (2010) PLX4032 in metastatic colorectal cancer J. Clin. Oncol. 28, 3555–3561
patients with mutant BRAF tumors. J. Clin. Oncol. 28, 3534 75. Larsen, M.J. et al. (2013) Classifications within molecular sub-
51. Kaufman, B. et al. (2015) Olaparib monotherapy in patients with types enables identification of BRCA1/BRCA2 mutation carriers
advanced cancer and a germline BRCA1/2 mutation. J. Clin. by RNA tumor profiling. PLoS One 8, e64268
Oncol. 33, 244–250 76. Van Allen, E.M. et al. (2015) Genomic correlates of response to
52. Al-Ahmadie, H. et al. (2014) Synthetic lethality in ATM-deficient CTLA–4 blockade in metastatic melanoma. Science 350,
RAD50-mutant tumors underlies outlier response to cancer 207–211
therapy. Cancer Discov. 4, 1014–1021 77. Hugo, W. et al. (2016) Genomic and transcriptomic features of
53. Banerji, U. et al. (2014) Abstract LB-66: results of two phase I response to anti-PD-1 therapy in metastatic melanoma. Cell
multicenter trials of AZD5363, an inhibitor of AKT1, 2 and 3: 165, 35–44

78. Ewing, A.D. et al. (2015) Combining tumor genome simulation 103. Ciriello, G. et al. (2012) Mutual exclusivity analysis identifies
with crowdsourcing to benchmark somatic single-nucleotide- oncogenic network modules. Genome Res. 22, 398–406
variant detection. Nat. Methods 12, 623–630 104. Kim, Y.A. et al. (2017) WeSME: uncovering mutual exclusivity of
79. Zehir, A. et al. (2017) Mutational landscape of metastatic cancer cancer drivers and beyond. Bioinformatics 33, 814–821
revealed from prospective clinical sequencing of 10,000 105. Leiserson, M.D. et al. (2016) A weighted exact test for mutually
patients. Nat. Med. 23, 703–713 exclusive mutations in cancer. Bioinformatics 32, i736–i745
80. Ciriello, G. et al. (2013) Emerging landscape of oncogenic sig- 106. Miller, C.A. et al. (2011) Discovering functional modules by
natures across human cancers. Nat. Genet. 45, 1127–1133 identifying recurrent and mutually exclusive mutational patterns
81. Wang, B. et al. (2014) Similarity network fusion for aggregating in tumors. BMC Med. Genomics 4, 34
data types on a genomic scale. Nat. Methods 11, 333–337 107. Vandin, F. et al. (2012) De novo discovery of mutated driver
82. Meng, C. et al. (2016) Dimension reduction techniques for the pathways in cancer. Genome Res. 22, 375–385
integrative analysis of multi-omics data. Brief. Bioinform. 17, 108. Vandin, F. et al. (2016) On the sample complexity of cancer
628–641 pathways identification. J. Comput. Biol. 23, 30–41
83. Way, G.P. and Greene, C.S. (2017) Extracting a biologically 109. Huang, F.W. et al. (2013) Highly recurrent TERT promoter muta-
relevant latent space from cancer transcriptomes with variations in human melanoma. Science 339, 957–959
tional autoencoders. bioRxiv Published online August 11,
110. Ainscough, B.J. et al. (2016) DoCM: a database of curated
2017. http://dx.doi.org/10.1101/174474
mutations in cancer. Nat. Methods 13, 806–807
84. Ching, T. et al. (2017) Opportunities and obstacles for deep
111. Griffith, M. et al. (2017) CIViC is a community knowledgebase for
learning in biology and medicine. bioRxiv Published online May
expert crowdsourcing the clinical interpretation of variants in
28, 2017. http://dx.doi.org/10.1101/142760
cancer. Nat. Genet. 49, 170–174
85. Curtis, C. et al. (2012) The genomic and transcriptomic archi-
112. Leiserson, M.D. et al. (2015) MAGI: visualization and collabora-
tecture of 2,000 breast tumours reveals novel subgroups.
tive annotation of genomic aberrations. Nat. Methods 12,
Nature 486, 346–352
483–484
86. Dienstmann, R. et al. (2017) Consensus molecular subtypes and
113. Gerstung, M. et al. (2017) Precision oncology for acute myeloid
the evolution of precision medicine in colorectal cancer. Nat.
leukemia using a knowledge bank approach. Nat. Genet. 49,
Rev. Cancer 17, 79–92
332–340
87. Le, D.T. et al. (2017) Mismatch repair deficiency predicts
114. Brough, R. et al. (2011) Searching for synthetic lethality in
response of solid tumors to PD-1 blockade. Science 357,
cancer. Curr. Opin. Genet. Dev. 21, 34–41
409–413
115. Hartwell, L.H. et al. (1997) Integrating genetic approaches into
88. Ng, P.C. and Henikoff, S. (2003) SIFT: predicting amino acid
the discovery of anticancer drugs. Science 278, 1064–1068
changes that affect protein function. Nucleic Acids Res. 31,
3812–3814 116. Lord, C.J. et al. (2008) A high-throughput RNA interference
screen for DNA repair determinants of PARP inhibitor sensitivity.
89. Reva, B. et al. (2011) Predicting the functional impact of protein
DNA Repair (Amst) 7, 2010–2019
mutations: application to cancer genomics. Nucleic Acids Res.
39, e118 117. Luo, J. et al. (2009) A genome-wide RNAi screen identifies
multiple synthetic lethal interactions with the Ras oncogene.
90. Chang, M.T. et al. (2016) Identifying recurrent mutations in
Cell 137, 835–848
cancer reveals widespread lineage diversity and mutational
specificity. Nat. Biotechnol. 34, 155–163 118. Martin, S.A. et al. (2009) Methotrexate induces oxidative DNA
damage and is selectively lethal to tumour cells with defects in
91. Gao, J. et al. (2017) 3D clusters of somatic mutations in cancer
the DNA mismatch repair gene MSH2. EMBO Mol. Med. 1,
reveal numerous rare mutations as functional targets. Genome
323–337
Med. 9, 4
119. Srivas, R. et al. (2016) A network of conserved synthetic lethal
92. Miller, M.L. et al. (2015) Pan-cancer analysis of mutation hot-
interactions for exploration of precision cancer therapy. Mol. Cell
spots in protein domains. Cell Syst. 1, 197–209
63, 514–525
93. Jia, P. and Zhao, Z. (2014) VarWalker: personalized mutation
120. Wang, T. et al. (2017) Gene essentiality profiling reveals gene
network analysis of putative cancer genes from next-generation
networks and synthetic lethal interactions with oncogenic ras.
sequencing data. PLoS Comput. Biol. 10, e1003460
Cell 168, 890–903
94. Lawrence, M.S. et al. (2013) Mutational heterogeneity in cancer
121. Wang, X. et al. (2014) Widespread genetic epistasis among
and the search for new cancer-associated genes. Nature 499,
cancer genes. Nat. Commun. 5, 4828
214–218
122. Aguirre, A.J. et al. (2016) Genomic copy number dictates a
95. Polak, P. et al. (2015) Cell-of-origin chromatin organization
gene-independent cell response to CRISPR/Cas9 targeting.
shapes the mutational landscape of cancer. Nature 518,
Cancer Discov. 6, 914–929
360–364
123. Cheung, H.W. et al. (2011) Systematic investigation of genetic
96. Mermel, C.H. et al. (2011) GISTIC2.0 facilitates sensitive and
vulnerabilities across cancer cell lines reveals lineage-specific
confident localization of the targets of focal somatic copy-num-
dependencies in ovarian cancer. Proc. Natl. Acad. Sci. U. S. A.
ber alteration in human cancers. Genome Biol. 12, R41
108, 12372–12377
97. Wendl, M.C. et al. (2011) PathScan: a tool for discerning muta-
124. Cowley, G.S. et al. (2014) Parallel genome-scale loss of function
tional significance in groups of putative cancer genes. Bioinfor-
screens in 216 cancer cell lines for the identification of context-
matics 27, 1595–1602
specific genetic dependencies. Sci. Data 1, 140035
98. Glaab, E. et al. (2012) EnrichNet: network-based gene set
125. Marcotte, R. et al. (2012) Essential gene profiles in breast,
enrichment analysis. Bioinformatics 28, i451–i457
pancreatic, and ovarian cancer cells. Cancer Discov. 2,
99. Vandin, F. et al. (2011) Algorithms for detecting significantly 172–189
mutated pathways in cancer. J. Comput. Biol. 18, 507–522
126. Marcotte, R. et al. (2016) Functional genomic landscape of
100. Vanunu, O. et al. (2010) Associating genes and protein com- human breast cancer drivers, vulnerabilities, and resistance.
plexes with disease via network propagation. PLoS Comput. Cell 164, 293–309
Biol. 6, e1000641
127. Beijersbergen, R.L. et al. (2017) Synthetic lethality in cancer
101. Kim, J.W. et al. (2016) Characterizing genomic alterations in therapeutics. Ann. Rev. Cancer Biol. 1, 141–161
cancer by complementary functional associations. Nat. Biotech-
128. Costanzo, M. et al. (2010) The genetic landscape of a cell.
nol. 34, 539–546
Science 327, 425–431
102. Knijnenburg, T.A. et al. (2016) Logic models to predict continu-
129. Lu, X. et al. (2013) Genome evolution predicts genetic interac-
ous outputs based on binary inputs with an application to
tions in protein complexes and reveals cancer drug targets. Nat.
personalized cancer therapy. Sci. Rep. 6, 36812
Commun. 4, 2124

130. Madhukar, N.S. et al. (2015) Prediction of genetic interactions 157. Hubert, C.G. et al. (2016) A three-dimensional organoid culture
using machine learning and network properties. Front. Bioeng. system derived from human glioblastomas recapitulates the
Biotechnol. 3, 172 hypoxic gradients and cancer stem cell heterogeneity of tumors
131. Zhong, W. and Sternberg, P.W. (2006) Genome-wide prediction found in vivo. Cancer Res. 76, 2465–2477
of C elegans genetic interactions. Science 311, 1481–1484 158. Pauli, C. et al. (2017) Personalized in vitro and in vivo cancer
132. Conde-Pueyo, N. et al. (2009) Human synthetic lethal inference models to guide precision medicine. Cancer Discov. 7, 462–477
as potential anti-cancer target gene detection. BMC Syst. Biol. 159. Ercan, D. et al. (2015) EGFR mutations and resistance to irre-
3, 116 versible pyrimidine–based EGFR inhibitors. Clin. Cancer Res.
133. Folger, O. et al. (2011) Predicting selective drug targets in cancer 21, 3913–3923
through metabolic networks. Mol. Syst. Biol. 7, 501 160. Janne, P.A. et al. (2015) AZD9291 in EGFR inhibitor-resistant
134. Frezza, C. et al. (2011) Inborn and acquired metabolic defects in non-small-cell lung cancer. N. Engl. J. Med. 372, 1689–1699
cancer. J. Mol. Med. (Berl) 89, 213–220 161. Jia, Y. et al. (2016) Overcoming EGFR(T790M) and EGFR
135. Wang, K. et al. (2009) Genome-wide identification of post- (C797S) resistance with mutant-selective allosteric inhibitors.
translational modulators of transcription factor activity in human Nature 534, 129–132
B cells. Nat. Biotechnol. 27, 829–839 162. Sequist, L.V. et al. (2015) Rociletinib in EGFR-mutated non-
136. Jerby-Arnon, L. et al. (2014) Predicting cancer-specific vulnera- small-cell lung cancer. N. Engl. J. Med. 372, 1700–1709
bility via data-driven detection of synthetic lethality. Cell 158, 163. Zhou, W. et al. (2009) Novel mutant-selective EGFR kinase
1199–1209 inhibitors against EGFR T790M. Nature 462, 1070–1074
137. Park, S. and Lehner, B. (2015) Cancer type-dependent genetic 164. Bardelli, A. et al. (2013) Amplification of the MET receptor drives
interactions between cancer driver alterations indicate plasticity resistance to anti-EGFR therapies in colorectal cancer. Cancer
of epistasis across cell types. Mol. Syst. Biol. 11, 824 Discov. 3, 658–673
138. Szczurek, E. et al. (2013) Synthetic sickness or lethality points at 165. Bozic, I. et al. (2013) Evolutionary dynamics of cancer in
candidate combination therapy targets in glioblastoma. Int. J. response to targeted combination therapy. Elife 2, e00747
Cancer 133, 2123–2132 166. Shahbazian, D. et al. (2012) Vertical pathway targeting in cancer
139. Sinha, S. et al. (2017) Systematic discovery of mutation-specific therapy. Adv. Pharmacol. 65, 1–26
synthetic lethals by mining pan-cancer human primary tumor 167. Russo, M. et al. (2016) Tumor heterogeneity and lesion-specific
data. Nat. Commun. 8, 15580 response to targeted therapy in colorectal cancer. Cancer Dis-
140. Berger, A.H. et al. (2016) High-throughput phenotyping of lung cov. 6, 147–153
cancer somatic mutations. Cancer Cell 30, 214–228 168. Siravegna, G. et al. (2015) Clonal evolution and resistance to
141. Iorio, F. et al. (2016) A landscape of pharmacogenomic inter- EGFR blockade in the blood of colorectal cancer patients. Nat.
actions in cancer. Cell 166, 740–754 Med. 21, 795–801
142. Gao, H. et al. (2015) High-throughput screening using patient- 169. Das Thakur, M. et al. (2013) Modelling vemurafenib resistance in
derived tumor xenografts to predict clinical trial drug response. melanoma reveals a strategy to forestall drug resistance. Nature
Nat. Med. 21, 1318–1325 494, 251–255
143. Nagel, Z.D. et al. (2017) DNA repair capacity in multiple path- 170. Sun, C. et al. (2014) Reversible and adaptive resistance to BRAF
ways predicts chemoresistance in glioblastoma multiforme. (V600E) inhibition in melanoma. Nature 508, 118–122
Cancer Res. 77, 198–206 171. Enriquez-Navas, P.M. et al. (2016) Exploiting evolutionary prin-
144. Reinhold, W.C. et al. (2017) DNA-targeted precision medicine; ciples to prolong tumor control in preclinical models of breast
have we been caught sleeping? Trends Cancer 3, 2–6 cancer. Sci. Transl. Med. 8, 327ra24
145. Friedman, A.A. et al. (2015) Precision medicine for cancer with 172. Diaz, L.A., Jr et al. (2012) The molecular evolution of acquired
next-generation functional diagnostics. Nat. Rev. Cancer 15, resistance to targeted EGFR blockade in colorectal cancers.
747–756 Nature 486, 537–540
146. Byrne, A.T. et al. (2017) Interrogating open issues in cancer 173. Girotti, M.R. et al. (2016) Application of sequencing, liquid biop-
precision medicine with patient-derived xenografts. Nat. Rev. sies, and patient-derived xenografts for personalized medicine
Cancer 17, 254–268 in melanoma. Cancer Discov. 6, 286–299
147. Garralda, E. et al. (2014) Integrated next-generation sequencing 174. Misale, S. et al. (2012) Emergence of KRAS mutations and
and avatar mouse models for personalized cancer treatment. acquired resistance to anti-EGFR therapy in colorectal cancer.
Clin. Cancer Res. 20, 2476–2484 Nature 486, 532–536
148. Hidalgo, M. et al. (2011) A pilot clinical study of treatment guided 175. Murtaza, M. et al. (2013) Non-invasive analysis of acquired
by personalized tumorgrafts in patients with advanced cancer. resistance to cancer therapy by sequencing of plasma DNA.
Mol. Cancer Ther. 10, 1311–1316 Nature 497, 108–112
149. Stebbing, J. et al. (2014) Patient-derived xenografts for individ- 176. Kolch, W. et al. (2015) The dynamic control of signal transduc-
ualized care in advanced sarcoma. Cancer 120, 2006–2015 tion networks in cancer cells. Nat. Rev. Cancer 15, 515–527
150. Fatehullah, A. et al. (2016) Organoids as an in vitro model of 177. Konieczkowski, D.J. et al. (2014) A melanoma cell state distinc-
human development and disease. Nat. Cell Biol. 18, 246–254 tion influences sensitivity to MAPK pathway inhibitors. Cancer
151. Boj, S.F. et al. (2015) Organoid models of human and mouse Discov. 4, 816–827
ductal pancreatic cancer. Cell 160, 324–338 178. Muller, J. et al. (2014) Low MITF/AXL ratio predicts early resistance
152. Sato, T. et al. (2011) Long-term expansion of epithelial organo- to multiple targeted drugs in melanoma. Nat. Commun. 5, 5712
ids from human colon, adenoma, adenocarcinoma, and Bar- 179. Tirosh, I. et al. (2016) Dissecting the multicellular ecosystem of
rett’s epithelium. Gastroenterology 141, 1762–1772 metastatic melanoma by single-cell RNA-seq. Science 352,
153. van de Wetering, M. et al. (2015) Prospective derivation of a 189–196
living organoid biobank of colorectal cancer patients. Cell 161, 180. Jordan, N.V. et al. (2016) HER2 expression identifies dynamic
933–945 functional states within circulating breast cancer cells. Nature
154. Weeber, F. et al. (2015) Preserved genetic diversity in organoids 537, 102–106
cultured from biopsies of human colorectal cancer metastases. 181. Nieto, M.A. et al. (2016) EMT: 2016. Cell 166, 21–45
Proc. Natl. Acad. Sci. U. S. A. 112, 13308–13311 182. Fischer, K.R. et al. (2015) Epithelial-to-mesenchymal transition is
155. Nadauld, L.D. et al. (2014) Metastatic tumor evolution and not required for lung metastasis but contributes to chemore-
organoid modeling implicate TGFBR2 as a cancer driver in sistance. Nature 527, 472–426
diffuse gastric cancer. Genome Biol. 15, 428 183. Zheng, X. et al. (2015) Epithelial-to-mesenchymal transition is
156. Gao, D. et al. (2014) Organoid cultures derived from patients dispensable for metastasis but induces chemoresistance in
with advanced prostate cancer. Cell 159, 176–187 pancreatic cancer. Nature 527, 525–530

184. Sequist, L.V. et al. (2011) Genotypic and histological evolution of 210. Liu, X. et al. (2017) Conditional reprogramming and long-term
lung cancers acquiring resistance to EGFR inhibitors. Sci. expansion of normal and tumor cells from human biospecimens.
Transl. Med. 3, 75ra26 Nat. Protoc. 12, 439–451
185. Dey, N. et al. (2015) Evading anti-angiogenic therapy: resistance 211. Beglyarova, N. et al. (2016) Screening of conditionally reprog-
to anti-angiogenic therapy in solid tumors. Am. J. Transl. Res. 7, rammed patient-derived carcinoma cells identifies ERCC3-MYC
1675–1698 interactions as a target in pancreatic cancer. Clin. Cancer Res.
186. Hugo, W. et al. (2015) Non-genomic and immune evolution of 22, 6153–6163
melanoma acquiring MAPKi resistance. Cell 162, 1271–1285 212. Saeed, K. et al. (2017) Comprehensive drug testing of patient-
187. Alix-Panabieres, C. and Pantel, K. (2016) Clinical applications of derived conditionally reprogrammed cells from castration-resis-
circulating tumor cells and circulating tumor DNA as liquid tant prostate cancer. Eur. Urol. 71, 319–327
biopsy. Cancer Discov. 6, 479–491 213. Crystal, A.S. et al. (2014) Patient-derived models of acquired
188. Fong, C.Y. et al. (2015) BET inhibitor resistance emerges from resistance can identify effective drug combinations for cancer.
leukaemia stem cells. Nature 525, 538–542 Science 346, 1480–1486
189. Miyamoto, D.T. et al. (2015) RNA-Seq of single prostate CTCs 214. Ridky, T.W. et al. (2010) Invasive three-dimensional organotypic
implicates noncanonical Wnt signaling in antiandrogen resis- neoplasia from multiple normal human epithelia. Nat. Med. 16,
tance. Science 349, 1351–1356 1450–1455
190. Motter, A.E. et al. (2008) Predicting synthetic rescues in meta- 215. Kenny, H.A. et al. (2015) Quantitative high throughput screening
bolic networks. Mol. Syst. Biol. 4, 168 using a primary human three-dimensional organotypic culture
predicts in vivo efficacy. Nat. Commun. 6, 6220
191. Rathert, P. et al. (2015) Transcriptional plasticity promotes pri-
mary and acquired resistance to BET inhibition. Nature 525, 216. Vaira, V. et al. (2010) Preclinical model of organotypic culture for
543–547 pharmacodynamic profiling of human tumors. Proc. Natl. Acad.
Sci. U. S. A. 107, 8352–8356
192. van Leeuwen, J. et al. (2016) Exploring genetic suppression
interactions on a global scale. Science 354, aag0839 217. Freeman, A.E. and Hoffman, R.M. (1986) In vivo-like growth of
human tumors in vitro. Proc. Natl. Acad. Sci. U. S. A. 83,
193. Tsimberidou, A.M. et al. (2012) Personalized medicine in a
2694–2698
phase I clinical trials program: the MD Anderson Cancer Center
initiative. Clin. Cancer Res. 18, 6373–6383 218. Meijer, T.G. et al. (2017) Ex vivo tumor culture systems for
functional drug testing and therapy response prediction. Future
194. Schwaederle, M. et al. (2016) Precision oncology: the UC San
Sci. OA 3, FSO190
Diego Moores Cancer Center PREDICT experience. Mol. Can-
cer Ther. 15, 743–752 219. Majumder, B. et al. (2015) Predicting clinical response to anti-
cancer drugs using anex vivo platform that captures tumour
195. Von Hoff, D.D. et al. (2010) Pilot study using molecular profiling
heterogeneity. Nat. Commun. 6, 6169
of patients’ tumors to find potential targets and select treat-
ments for their refractory cancers. J. Clin. Oncol. 28, 4877–4883 220. Holliday, D.L. et al. (2013) The practicalities of using tissue slices
as preclinical organotypic breast cancer models. J. Clin. Pathol.
196. Park, J.W. et al. (2016) Adaptive randomization of neratinib in
66, 253–255
early breast cancer. N. Engl. J. Med. 375, 11–22
221. Naipal, K.A. et al. (2016) Tumor slice culture system to assess
197. Rugo, H.S. et al. (2016) Adaptive randomization of veliparib-
drug response of primary breast cancer. BMC Cancer 16, 78
carboplatin treatment in breast cancer. N. Engl. J. Med. 375,
23–34 222. Roife, D. et al. (2016) Ex vivo testing of patient-derived xeno-
grafts mirrors the clinical outcome of patients with pancreatic
198. Tripathy, D. et al. (2015) Adaptively randomized trial of neo-
ductal adenocarcinoma. Clin. Cancer Res. 22, 6021–6030
adjuvant chemotherapy with or without the Akt inhibitor MK-
2206: graduation results from the I-SPY 2 Trial. J. Clin. Oncol. 223. Wang, Y. et al. (2014) Clonal evolution in breast cancer revealed
33, 524–524 by single nucleus genome sequencing. Nature 512, 155–160
199. Kaplan, R. et al. (2013) Evaluating many treatments and bio- 224. Patel, A.P. et al. (2014) Single-cell RNA-seq highlights intra-
markers in oncology: a new design. J. Clin. Oncol. 31, tumoral heterogeneity in primary glioblastoma. Science 344,
4562–4568 1396–1401
200. Liu, F. et al. (2015) EGFR mutation promotes glioblastoma 225. Tirosh, I. et al. (2016) Single-cell RNA-seq supports a develop-
through epigenome and transcription factor network remodel- mental hierarchy in human oligodendroglioma. Nature 539,
ing. Mol. Cell 60, 307–318 309–313
201. Zhu, J. et al. (2015) Gain-of-function p53 mutants co-opt chro- 226. Gawad, C. et al. (2016) Single-cell genome sequencing: current
matin pathways to drive cancer growth. Nature 525, 206–211 state of the science. Nat. Rev. Genet. 17, 175–188
202. Mertins, P. et al. (2016) Proteogenomics connects somatic 227. Vitak, S.A. et al. (2017) Sequencing thousands of single-cell
mutations to signalling in breast cancer. Nature 534, 55–62 genomes with combinatorial indexing. Nat. Methods 14,
302–308
203. Zhang, B. et al. (2014) Proteogenomic characterization of
human colon and rectal cancer. Nature 513, 382–387 228. Zahn, H. et al. (2017) Scalable whole-genome single-cell library
preparation without preamplification. Nat. Methods 14,
204. Zhang, H. et al. (2016) Integrated proteogenomic characteriza-
167–173
tion of human high-grade serous ovarian cancer. Cell 166,
755–765 229. Klein, A.M. et al. (2015) Droplet barcoding for single-cell tran-
scriptomics applied to embryonic stem cells. Cell 161,
205. Drake, J.M. et al. (2016) Phosphoproteome integration reveals
1187–1201
patient-specific networks in prostate cancer. Cell 166,
1041–1054 230. Macosko, E.Z. et al. (2015) Highly parallel genome-wide expres-
sion profiling of individual cells using nanoliter droplets. Cell 161,
206. Aebersold, R. and Mann, M. (2016) Mass-spectrometric explo-
1202–1214
ration of proteome structure and function. Nature 537, 347–355
231. Zheng, G.X. et al. (2017) Massively parallel digital transcriptional
207. Senft, D. and Ronai, Z.A. (2016) Adaptive stress responses
profiling of single cells. Nat. Commun. 8, 14049
during tumor metastasis and dormancy. Trends Cancer 2,
429–442 232. Tsoucas, D. and Yuan, G.C. (2017) Recent progress in single-
cell cancer genomics. Curr. Opin. Genet. Dev. 42, 22–32
208. Davidson, S.M. et al. (2016) Environment impacts the metabolic
dependencies of ras-driven non-small cell lung cancer. Cell
Metab. 23, 517–528
209. Hensley, C.T. et al. (2016) Metabolic heterogeneity in human
lung tumors. Cell 164, 681–694

PIIS147149991417301430

Uploaded by

Copyright:

Available Formats

PIIS147149991417301430

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PIIS147149991417301430

Uploaded by

Copyright:

Available Formats

Feature Review

Precision Oncology: The

Available patient-derived functional

Beyond direct targeting of genomic alterations, the impact of differentiation hierarchies,

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 875

(See figure legend on the bottom of the next page.)

876 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 877

878 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 879

880 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Approaches for Tumor Subclassiﬁcation

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 881

PREDICT [194]; All NGS; 347 87 N.S. More patients with SD

882 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Hotspot-seq BRAF, PIK3CA, KRAS, ORR, PFS, OS

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 883

FOCUS 4 [199]; Phase Advanced/metastatic, BRAFi + panitumumab +/–

884 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

The observed intertumor mutational heterogeneity is widely believed to be due, in part, to

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 885

Identifying Cancer Vulnerabilities on the Basis of Genetic Interactions

886 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 887

Box 2. Valuable Models for Guiding Precision Therapy

888 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Clinical Management of Therapy Resistance in the Precision Oncology Era

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 889

890 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

In addition to the Darwinian-like evolution of genomic alterations in resistant clones under

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 891

Concluding Remarks: The Road Ahead

892 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Box 3. Clinicians Corner

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 893

894 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 895

896 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

Trends in Molecular Medicine, October 2017, Vol. 23, No. 10 897

898 Trends in Molecular Medicine, October 2017, Vol. 23, No. 10

You might also like