McrobioogyML DL 2023
McrobioogyML DL 2023
McrobioogyML DL 2023
Advances in machine
OPEN ACCESS learning-based bacteria analysis
for forensic identification: identity,
EDITED BY
Chen Li,
Northeastern University, China
REVIEWED BY
Hua Cao,
ethnicity, and site of occurrence
Shanghai Jiao Tong University, China
Tongyun Liu,
Kunming Medical University, China
Geyao Xu 1,2, Xianzhuo Teng 3, Xing-Hua Gao 1,2, Li Zhang 1,2*,
Zhe Xu, Hongwei Yan 1,2* and Rui-Qun Qi 1,2*
Capital Medical University, China
1
Department of Dermatology, The First Hospital of China Medical University, Shenyang, China, 2 Key
*CORRESPONDENCE
Laboratory of Immunodermatology, Ministry of Education and NHC, National Joint Engineering
Rui-Qun Qi
Research Center for Theranostics of Immunological Skin Diseases, Shenyang, China, 3 Department of
[email protected]
Cardiology, Shengjing Hospital of China Medical University, Shenyang, China
Li Zhang
[email protected]
Hongwei Yan
[email protected] When faced with an unidentified body, identifying the victim can be challenging,
particularly if physical characteristics are obscured or masked. In recent years,
RECEIVED 03 November 2023
ACCEPTED 05 December 2023 microbiological analysis in forensic science has emerged as a cutting-edge
PUBLISHED 21 December 2023 technology. It not only exhibits individual specificity, distinguishing different
CITATION human biotraces from various sites of occurrence (e.g., gastrointestinal, oral, skin,
Xu G, Teng X, Gao X-H, Zhang L, Yan H and respiratory, and genitourinary tracts), each hosting distinct bacterial species, but
Qi R-Q (2023) Advances in machine learning-
based bacteria analysis for forensic
also offers insights into the accident’s location and the surrounding environment.
identification: identity, ethnicity, and site of The integration of machine learning with microbiomics provides a substantial
occurrence. improvement in classifying bacterial species compares to traditional sequencing
Front. Microbiol. 14:1332857.
doi: 10.3389/fmicb.2023.1332857
techniques. This review discusses the use of machine learning algorithms such
as RF, SVM, ANN, DNN, regression, and BN for the detection and identification of
COPYRIGHT
© 2023 Xu, Teng, Gao, Zhang, Yan and Qi. This various bacteria, including Bacillus anthracis, Acetobacter aceti, Staphylococcus
is an open-access article distributed under the aureus, and Streptococcus, among others. Deep leaning techniques, such as
terms of the Creative Commons Attribution Convolutional Neural Networks (CNN) models and derivatives, are also employed
License (CC BY). The use, distribution or
reproduction in other forums is permitted, to predict the victim’s age, gender, lifestyle, and racial characteristics. It is
provided the original author(s) and the anticipated that big data analytics and artificial intelligence will play a pivotal role
copyright owner(s) are credited and that the in advancing forensic microbiology in the future.
original publication in this journal is cited, in
accordance with accepted academic practice.
No use, distribution or reproduction is KEYWORDS
permitted which does not comply with these
terms. machine learning, deep leaning, artificial intelligence, forensic identification,
microbiological
1 Introduction
Microorganisms, encompassing fungi, bacteria and viruses, pervade both individuals and
their surroundings, existing ubiquitously in nature. The exploration of the human microbiome,
comprising microorganisms residing in, on, and around humans, has revolutionized our
comprehension of the intricate interactions between these microorganisms and human health
and disease (Johnson et al., 2016). The interplay of genetics and environment gives rise to a
distinct microbiome for each individual. The skin microbiome is continuously shed and
transferred from the host. Utilizing deposited microbes as supplementary evidence can aid in
either including or excluding individuals in a criminal case (Sherier et al., 2022). Previous
investigations have utilized microbial DNA markers, such as 16 s ribosomal RNA (rRNA), to
evaluate the health of human hosts. However, this method is constrained to contaminated
samples and hinges on the initial PCR amplification of labeled genes immune system typically prevents the colonization of internal organs
(Javan et al., 2016), limiting its applicability. The human microbiome and body fluids by microbes. However, after death, as the immune
presents a promising alternative for utilizing additional DNA sources system and physical barriers break down, microorganisms proliferate
in identifying or excluding individuals linked to erroneous evidence, throughout the body, starting in the gastrointestinal tract (Spagnolo
thereby enhancing the efficiency of forensic DNA analysis in criminal et al., 2019). Consequently, the human post-mortem microbiome has
investigations (Sherier et al., 2021). been studied and identified as consisting of two components:
Studies on the human microbiome have unveiled substantial microorganisms found in internal organs and body fluids after death,
variations in the composition and abundance of microbial and microorganisms located on the surface of the remains
communities across different body sites, varying states of host health, postmortem, known as epinecrotic (Benbow et al., 2015). Research
and among diverse racial groups. Notably, the skin flora is dominated indicates that bacteria on the external surface of the body hold
by Propionibacterium, Corynebacterium and Staphylococcus, while the potential as markers for forensic identification. Skin bacteria, being
oral flora is characterized by the prevalence of Lactobacillus and highly resistant to environmental changes such as ultraviolet radiation,
Haemophilus spp., and the gut flora is marked by the dominance of humidity and temperature, could serve as distinctive molecular
Clostridium and Mycobacterium spp. (Lei et al., 2022). For instance, “fingerprints” for humans (Javan et al., 2016). Furthermore, microbial
previous investigations have highlighted significant differences in the decomposition processes play a crucial role in postmortem changes.
microbiological composition of the hand microbiome alone, varying While the succession of microbial communities is generally consistent
across countries and regions. In the hand microbiota, across various soil types, the microbial community in the environment
Propionibacterium and Streptococcus were prominent in Americans, influences the decomposition processes of the cadaveric microbial
but Propionibacterium was largely absent in the microbiota of Koreans community. As decomposing carcasses release a range of substances,
and Japanese. Canadians, on the other hand, exhibited a hand including fatty tissues, volatile fatty acids, organic acids, organic
microbiota mainly composed of Bacillus, Streptococcus and nitrogen and specific anaerobic bacteria, into the soil (Wang et al.,
Propionibacterium (Cho and Eom, 2021). 2022), forensic identification based on microbial communities
In recent years, the field of forensic microbiology has gained becomes feasible. Nevertheless, microbial communities alone are
prominence in response to bioterrorist attacks. The primary objective susceptible to external factors that limit their utility in forensic
of microbiological forensics is to employ diverse methods, including identification-a challenge that artificial intelligence (AI) may address
microbiology, molecular biology, immunology and analytical by constructing effective assessment models.
chemistry, to deduce potential mutant strains of microorganisms Presently, machine learning (ML) stands as the primary AI
associated with bioterrorist attacks or natural disease outbreaks. It also technique employed in forensic research. ML, a subset of AI, devises
aims to predict microbial origins, affinities or routes of transmission algorithms to empower computers with the capability to learn without
(refer to Figure 1). The changes in microorganisms following the death explicit programming (Handelman et al., 2018), as illustrated in
of the host remain poorly understood. In healthy individuals, the Figure 2. Various ML methods, including Random Forest (RF),
FIGURE 1
Forensic microbiological processes.
FIGURE 2
Artificial intelligence process.
Support Vector Machines (SVM), Linear Regression, Logistic tissue or fluid samples. AI facilitates the comparison of field data with
Regression (LR), among others, play a crucial role in this context. One machine-generated data. By inputting an individual’s specific
notable subfield of ML is Deep Learning (DL), which encompasses biometric pattern into the system, the machine swiftly matches it with
Artificial Neural Networks (ANN), Multi-Layer Perceptron (MLP) pre-recorded biometric data to ascertain the individual’s identity
networks, and Convolutional Neural Networks (CNN). The efficiently (Wankhade et al., 2022).
intersection of AI and microbiomics holds promise for a more Sherier et al. (2021) conducted a study where subjects’
profound understanding of the role played by microbial communities non-dominant hands were sampled for skin microbes. They utilized
in cadavers. While conventional statistical methods are limited to the Hidskinplex multiple panels and SVM models to analyze markers
determining the broad composition of microbial communities and and predict donor identity. The Hidskinplex is an innovative targeted
their general successional changes, ML models facilitate quantitative sequencing panel comprising 286 skin core microbial markers,
analyses and accurate predictions (Yuan et al., 2023). This review encompassing a taxonomy of specific microorganisms highly
integrates various AI models with forensic science to investigate the abundant on human skin (Schmedes et al., 2018). The results
variability of microbial communities among different races and in demonstrated that the classification accuracy of single nucleotide
diverse geographic environments (such as seawater, freshwater, soil, polymorphisms (SNPs) using the chosen method reached up to
urban settings). The aim is to discern the individual identification, 94.77%. This conclusion was confirmed by the Schmedes et al. (2018).
gender, age, and health of a victim or suspect. Moreover, suspect The research indicates that the use of a RF learning model can not only
characterization can be conducted based on environmental identify individuals by combining individual-specific bacteria (such
microorganisms in the vicinity of the victim. as Cutibacterium acnes) with known genealogy data (Gu et al., 2022),
but also predict different body parts. Schmedes et al. used the RF
model to predict different body parts with an accuracy of 86%, while
2 Application of machine learning in the Tackmann team’s prediction rate reached 89% (Yang et al., 2019),
microbial forensic identification possibly because the Schmedes team only predicted three body parts:
shirt collar, touched items, and feet, while the Tackmann team
Traditional autopsies come with numerous limitations, requiring included a sample size of over 15,000. In 2022, Tackmann et al. (2018)
substantial manpower and often leading to differences in expert collected and examined genetic data from Chinese, Korean, and
opinions. AI emerges as a pivotal tool in various forensic science Japanese populations based on publicly available database. They
domains, addressing challenges in providing individual identity, utilized the Softmax and RF methods to filter information, ensuring
scrutinizing body stains or marks, and identifying and collecting the reliability of classification results with an accuracy exceeding 90%,
consistent with previous research findings. In summary, the use of anticipating Shigella, the most prevalent bacterium in Israel. The ML
various AI models allows for accurate identification of individual model demonstrated high accuracy in swiftly predicting the specific
information and specific body parts. However, future efforts should category of Shigella, aiding healthcare professionals in making real-
involve a larger sample size for predictive statistical analysis. time decisions about the need for hospitalization. Similarly, Njage
Subsequently, this elaboration delves into four key aspects of gender, et al. (2019) utilized the Logit Boost algorithm to predict the severity
age, health status and environmental cues—by amalgamting AI with of outcomes following infection with Shiga toxin-producing Escherichia
microbiome research. coli (E. coli), including symptoms such as diarrhea and hemolytic
uremic syndrome, achieving an accuracy rate of up to 75%. If these
models are integrated with whole-genome sequence data in the future,
2.1 Gender identification it could allow for the estimation of the health status of populations
affected by the infection.
In recent years, the application of AI models in gender In addition to using microbiota to assess an individual’s health, it
identification has proven to be valuable. A study found that the use of can also be employed to determine the cause of death in deceased
deep convolutional neural networks (DCNN) accurately predicted individuals. The practice of scanning the entire body of a decreased
gender across diverse chest X-ray datasets from different regions (Li individual using imaging methods, followed by computer software
et al., 2022). Gao et al. (2018) introduced a novel method, analysis to identify internal trauma and determine the actual cause of
MKDSIF-FCM (classification of skull datasets based on data mining death, is referred to as virtual autopsy. Sullivan et al. (2017) have
and data analysis), which relies on unsupervised classification proposed the development of virtual biospecimen repositories
techniques for gender identification of Han skulls. The algorithm enhanced with ML. The ML tool generates a series of non-invasive
demonstrated a notable 98 and 93.02% gender prediction accuracy for autopsy images utilizing X-rays, CT scans, or MRIs. Three images are
females and males, respectively, suggesting its potential in forensic then compared with the digital pathology data of corresponding
investigations. AI models not only demonstrate accurate predictive biospecimens to comprehend the disease process. This approach aims
value in determining human gender but also yield similar conclusions to assess clinical diagnoses and treatments, representing an emerging
in discerning the sex of dogs. Scarsella et al. (2020) employed a RF trend in the field. More research on virtual autopsy is still needed in
model to classify the microbial profiles of 340 healthy dogs based on the future, but this method is poised to become a new trend.
gender factors. The results revealed that, compared to intact dogs,
neutered males and spayed females formed distinct groups. These
findings indicate a potential bidirectional interaction between 2.4 Environmental identification
microbial communities and host endocrine states. In the future, the
integration of AI models based on microbial profiles holds greater The compositional complexity and diversity of soils make them
promise in gender identification. valuable for forensic identification. Soil pH, particularly in different
geographical locations, plays a crucial role in shaping bacterial
community (Gürtler et al., 2014). Ahmad et al. (2023) used various
2.2 Age identification ML methods to classify soil rich in Coxella bacteria. They were able to
identify correlations with elements such as clay, organic matter, trace
Wang et al. (2021) developed a novel DL prediction method based elements, and slit, achieving a classification accuracy of up to 84.35%.
on microbiome called MDeep. They applied this method to predict Utilizing ANN assisted by Fourier-transform infrared spectroscopy, a
the age of 531 Americans by analyzing their gut microbiota. The ML approach was applied to differentiate Bacillus cereus group
results indicated that the MDeep model exhibited higher accuracy in members in soil and foodborne outbreak samples. The study revealed
age prediction compared to other learning models, as evidenced by a that psychrotolerant B. cereus members in Bacillus anthracis, B. cereus,
significantly improved overall R-squared value. Moreover, the MDeep and B. weihenstephanensis dominated in soil samples from Germany,
model consistently demonstrated stability in the prediction (infant vs. Malta and Switzerland (53–70%), with Bacillus thuringiensis in the
child, child vs. adult) across sensitivity, specificity, accuracy, and minority (3–12%). In Danish soil samples, B. weihenstephanensis and
precision metrics. Additionally, the model could assess the host’s age B. mycoides were predominant (94%), with Bacillus anthracis
based on the maturity of the microbial community. Furthermore, accounted for only 6% (Bağcıoğlu et al., 2019). Consequently,
MDeep proved effective in determining host age based on colony examining soil microorganisms in the vicinity of a carcass may offer
maturation. Subramanian et al. (2014) combined a sparse model with insights into predicting the location of death.
a 16 s rRNA dataset, resulting in the accurate prediction of the relative Microorganisms are also utilized for environmental identification
maturity and age score of the microbiota. Although promising, more in marine and freshwater settings. Ahmed et al. (2018) employed the
research on microbiome-based AI for age prediction is still essential SourceTracker program (an ML tool based on high-throughput
in the future to further provide a more robust theoretical foundation sequencing data for microbial community tracking), studied the decay
for forensic identification. of sewage-associated bacterial communities in water. They found an
increase in the abundance of cells from Flavobacteriaceae and
Spirochaetaceae families in seawater. Similarly, in another study using
2.3 Health status identification the same program to detect bacteria in the middle and lower reaches
of rivers in Russia, it was observed that during winter and summer,
Adamker et al. (2018) employed three ML algorithms—LR, neural there was a significant increase in the abundance of E. coli and
networks (NN), and SVM—to construct predictive models for Enterococci; rod-shaped bacteria proliferated in sediments and other
environmental habitats, showing host specificity. Moreover, collection of data have not been universally established, and there is
Flavobacterium, Sphingobacterium, and Serratia were identified as the a need for large quantities of biometric data, despite challenges in
most common bacteria in freshwater ecosystems, with the first two infrastructure and resources availability. The diversity in AI methods
being abundant in debris particles and algal plants. β-Serratia emerged poses challenges in terms of standardization and interoperability, and
as the predominant bacterium in organic aggregates and in streams issues such as cross-contamination between samples and host DNA
with a high debris load (Dubinsky et al., 2016). contamination need to be addressed. Looking forward, with the
Furthermore, the use of microorganisms enables the expansion of and the accumulation of data from microbial images,
identification of living environments. In the daily environments with ML and even DL technologies are expected to wield even greater
which humans interact, approximately 360,000 micro-organisms are influence. The integration of AI and microbiomics holds promise in
released per hour, presenting a valuable resource for forensic enhancing investigative opinions and conclusions with higher
analysis. Furthermore, the human microbiome encompasses rare accuracy and timeliness, furnishing additional clues and a solid basis
microbial taxa, forming a unique microbial spectrum. Lax et al. for criminal cases investigators. Continued advancements in both
(2014) employed a Bayesian model to conduct a six-week monitoring fields and the resolution of current challenges will likely contribute
of microbial community changes in seven American households. to the further refinement and application of these technologies in
They discovered that the greatest differences in microbial species forensic science. In the future, machine learning models can facilitate
occurred in floor environments, while microbial populations on the integration of various species of microorganisms, applying them
doorknobs were the most similar. Additionally, associations were practically through processes such as detection, data transmission,
identified in the distribution of microorganisms across various and analysis.
surfaces in homes. In addition, the discovery of 4,728 novel bacterial
species over a span of three years in global urban public
transportation systems has expanded the diversity of known urban Author contributions
microbiomes. This findings enhances the analysis of microbial
interactions between humans and urban environments (Wu et al., GX: Investigation, Writing – original draft. XT: Writing – original
2022). Huang et al. (2020) employed LR and L2 normalization draft. X-HG: Writing – review & editing. LZ: Writing – review &
method to infer city affiliation in microbial samples, achieving an editing. HY: Writing – review & editing. R-QQ: Writing – review &
accuracy of up to 80%. Given the significant impact of urban editing, Conceptualization, Supervision.
microbial communities on human life, there is an ongoing need for
future research focusing on the intricate interrelationships between
human microbes and urban microbes. Funding
The author(s) declare that no financial support was received for
3 Discussion the research, authorship, and/or publication of this article.
References
Adamker, G., Holzer, T., Karakis, I., Amitay, M., Anis, E., Singer, S. R., et al. (2018). Ahmed, W., Staley, C., Kaiser, T., Sadowsky, M. J., Kozak, S., Beale, D., et al. (2018). Decay
Prediction of shigellosis outcomes in Israel using machine learning classifiers. Epidemiol. of sewage-associated bacterial communities in fresh and marine. Environmental waters and
Infect. 146, 1445–1451. doi: 10.1017/S0950268818001498 sediment. Appl. Microbiol. Biotechnol. 102, 7159–7170. doi: 10.1007/s00253-018-9112-4
Ahmad, F., Khan, M. U. G., Tahir, A., Tipu, M. Y., Rabbani, M., and Shabbir, M. Z. Bağcıoğlu, M., Fricker, M., Johler, S., and Ehling-Schulz, M. (2019). Detection and
(2023). Two phase feature-ranking for new soil dataset for Coxiella burnetii persistence identification of Bacillus cereus, Bacillus cytotoxicus, Bacillus thuringiensis, bacillus
and classification using machine learning models. Sci. Rep. 13, 022–26956. doi: 10.1038/ mycoides and Bacillus weihenstephanensis via machine learning based FTIR spectroscopy.
s41598-022-26956-8 Front. Microbiol. 10:902. doi: 10.3389/fmicb.2019.00902
Benbow, M. E., Pechal, J. L., Lang, J. M., Erb, R., and Wallace, J. R. (2015). The Scarsella, E., Stefanon, B., Cintio, M., Licastro, D., Sgorlon, S., Monego, S. D., et al.
potential of high-throughput metagenomic sequencing of aquatic bacterial communities (2020). Learning machine approach reveals microbial signatures of diet and sex in dog.
to estimate the postmortem submersion interval. J. Forensic Sci. 60, 1500–1510. doi: PLoS One 15:0237874. doi: 10.1371/journal.pone.0237874
10.1111/1556-4029.12859
Schmedes, S. E., Woerner, A. E., Novroski, N. M. M., Wendt, F. R., King, J. L.,
Cho, H.-W., and Eom, Y.-B. (2021). Forensic analysis of human microbiome in skin Stephens, K. M., et al. (2018). Targeted sequencing of clade-specific markers from skin
and body fluids based on geographic location. Front. Cell. Infect. Microbiol. 11:695191. microbiomes for forensic human identification. Forensic Sci. Int. Genet. 32, 50–61. doi:
doi: 10.3389/fcimb.2021.695191 10.1016/j.fsigen.2017.10.004
Dubinsky, E. A., Butkus, S. R., and Andersen, G. L. (2016). Microbial source tracking Sherier, A. J., Woerner, A. E., and Budowle, B. (2021). Population informative markers
in impaired watersheds using Phylo Chip and machine-learning classification. Water selected using Wright's fixation index and machine learning improves human
Res. 105, 56–64. doi: 10.1016/j.watres.2016.08.035 identification using the skin microbiome. Appl. Environ. Microbiome 87, 01208–01221.
doi: 10.1128/AEM.01208-21
Gao, H., Geng, G., and Yang, W. (2018). Sex determination of 3D skull based on a
novel unsupervised learning method. Comput. Math. Methods Med. 2018, 1–10. doi: Sherier, A. J., Woerner, A. E., and Budowle, B. (2022). Determining informative
10.1155/2018/4567267 microbial single nucleotide polymorphisms for human identification. Appl. Environ.
Microbiol. 88, 00052–00022. doi: 10.1128/aem.00052-22
Gu, J. Q., Zhao, H., Guo, X. Y., Sun, H. Y., Xu, J. Y., and Wei, Y. L. (2022). A high-
performance SNP panel developed by machine-learning approaches for characterizing Spagnolo, E. V., Stassi, C., Mondello, C., Zerbo, S., Milone, L., and Argo, A. (2019).
genetic differences of southern and northern Han Chinese, Korean, and Japanese Forensic microbiology applications: A systematic review. Legal Med. 36, 73–80. doi:
individuals. Electrophoresis 43, 1183–1192. doi: 10.1002/elps.202100184 10.1016/j.legalmed.2018.11.002
Gürtler, L., Bauerfeind, U., Blümel, J., Burger, R., Drosten, A., Gröner, A., et al. (2014). Subramanian, S., Huq, S., Yatsunenko, T., Haque, R., Mahfuz, M., Alam, M. A., et al.
Coxiella burnetii – pathogenic agent of Q (query) fever. Transfus. Med. Hemother. 41, (2014). Persistent gut microbiota immaturity in malnourished Bangladeshi children.
60–72. doi: 10.1159/000357107 Nature 510, 417–421. doi: 10.1038/nature13421
Handelman, G. S., Kok, H. K., Chandra, R. V., Razavi, A. H., Lee, M. J., and Asadi, H. Sullivan, S. O., Holzinger, A., Zatloukal, K., Saldiva, P., Sajid, M. I., and Wichmann, D.
(2018). eDoctor: machine learning and the future of medicine. J. Intern. Med. 284, (2017). Machine learning enhanced virtual autopsy. Autops. Case Rep. 7, 3–7. doi:
603–619. doi: 10.1111/joim.12822 10.4322/acr.2017.037
Huang, L., Xu, C., Yang, W., and Yu, R. (2020). A machine learning framework to Tackmann, J., Arora, N., Schmidt, T. S. B., Rodrigues, J. F. M., Mering, C., et al. (2018).
determine geolocations from metagenomic profiling. Biol. Direct 15, 020–00278. doi: Ecologically informed microbial biomarkers and accurate classification of mixed and
10.1186/s13062-020-00278-z unmixed samples in an extensive cross-study of human body sites. Microbiome 6,
018–0565. doi: 10.1186/s40168-018-0565-6
Javan, G. T., Finley, S. J., Abidin, Z., and Mulle, J. G. (2016). The Thanatomicrobiome:
a missing piece of the microbial puzzle of death. Front. Microbiol. 7:225. doi: 10.3389/ Wang, Y., Bhattacharya, T., Jiang, Y., Qin, X., Wang, Y., Liu, Y., et al. (2021). A novel
fmicb.2016.00225 deep learning method for predictive modeling of microbiome data. Brief. Bioinform.
22:73. doi: 10.1093/bib/bbaa073
Johnson, H. R., Trinidad, D. D., Guzman, S., Khan, Z., Parziale, J. V., DeBruyn, J. M.,
et al. (2016). A machine learning approach for using the postmortem skin microbiome Wang, Z., Zhang, F., Wang, L., Yuan, H., Guan, D., and Zhao, R. (2022). Advances in
to estimate. PLoS One 11:0167370. doi: 10.1371/journal.pone.0167370 artificial intelligence-based microbiome for PMI estimation. Front. Microbiol.
13:1034051. doi: 10.3389/fmicb.2022.1034051
Lax, S., Smith, D. P., Hampton-Marcell, J., Owens, S. M., Handley, K. M.,
Scott, N. M., et al. (2014). Longitudinal analysis of microbial interaction between Wankhade, T. D., Ingale, S. W., Mohite, P. M., and Bankar, N. J. (2022). Artificial
humans and the indoor environment. Science 345, 1048–1052. doi: 10.1126/ intelligence in forensic medicine and toxicology: the future of Forsenic medicine. Cureus
science.1254529 14:28376. doi: 10.7759/cureus.28376
Lei, F. Z., Chen, M., Mei, S. Y., Fang, Y. T., and Zhu, B. F. (2022). New advances, Wu, J., Danko, D., Afshinnekoo, E., Bezdan, D., Bhattacharyya, M., Castro-Nallar, E.,
challenges and opportunities in forensic applications of microbiomics. Fa Yi Xue Za Zhi et al. (2022). Annotating unknown species of urban microorganisms on a global scale
38, 625–639. doi: 10.12116/j.issn.1004-5619.2022.520303 unveils novel functional diversity and local environment association. Environ. Res.
207:112183. doi: 10.1016/j.envres.2021.112183
Li, D., Lin, C. T., Sulam, J., and Yi, P. H. (2022). Deep learning prediction of sex on
chest radiographs: a potential contributor to biased algorithms. Emerg. Radiol. 29, Yang, J., Tsukimi, T., Yoshikawa, M., Suzuki, K., Takeda, T., Tomita, M., et al. (2019).
365–370. doi: 10.1007/s10140-022-02019-3 Cutibacterium acnes (Propionibacterium acnes) 16S rRNA genotyping of microbial
samples from possessions contributes to owner identification. mSystems 4, 00594–00519.
Njage, P. M. K., Leekitcharoenphon, P., and Hald, T. (2019). Improving
doi: 10.1128/mSystems.00594-19
hazard characterization in microbial risk assessment using next generation
sequencing data and machine learning: predicting clinical outcomes in shigatoxigenic Yuan, H., Wang, Z., Wang, Z., Zhang, F., Guan, D., and Zhao, R. (2023). Trends in
Escherichia coli. Int. J. Food Microbiol. 292, 72–82. doi: 10.1016/j.ijfoodmicro. forensic microbiology: from classical methods to deep learning. Front. Microbiol.
2018.11.016 14:1163741. doi: 10.3389/fmicb.2023.1163741