AllPaths-LG Manual PDF
AllPaths-LG Manual PDF
AllPaths-LG Manual PDF
Computational Research and Development Group
Genome Sequencing and Analysis Program
Broad Institute of MIT and Harvard
Cambridge, MA
Manual Revision: (27‐Jan‐13 2:47:00 PM)
Table of Contents
ALLPATHSLG Manual .......................................................................................................................... 1
Conventions .................................................................................................................................................. 4
Introduction .................................................................................................................................................. 4
Capabilities and limitations ........................................................................................................................... 4
Staying up to date with our blog ................................................................................................................... 5
Requirements ................................................................................................................................................ 5
Availability ..................................................................................................................................................... 6
Getting Help .................................................................................................................................................. 6
Installation .................................................................................................................................................... 6
Troubleshooting ........................................................................................................................................ 7
Environment .............................................................................................................................................. 7
ALLPATHS pipeline overview ......................................................................................................................... 7
RunAllPathsLG module .............................................................................................................................. 7
ALLPATHS pipeline directory structure ...................................................................................................... 7
REFERENCE (organism) directory ....................................................................................................................... 8
DATA (project) directory ...................................................................................................................................... 8
RUN (assembly pre‐processing) directory ............................................................................................................. 8
ASSEMBLIES directory ....................................................................................................................................... 9
SUBDIR (assembly) directory .............................................................................................................................. 9
Required ALLPATHS arguments .................................................................................................................... 9
Preparing data for ALLPATHS ........................................................................................................................ 9
Supported library constructions ................................................................................................................ 9
Read orientation ..................................................................................................................................... 10
ALLPATHS input files................................................................................................................................ 10
Base, quality score, and pairing information files ............................................................................................... 10
The ploidy file ................................................................................................................................................. 11
Preparing ALLPATHS input files ............................................................................................................... 11
Accepted data file formats .................................................................................................................................. 11
in_groups.csv file ....................................................................................................................................... 11
in_libs.csv file ............................................................................................................................................ 12
Running conversion script ................................................................................................................................... 13
Import reference ..................................................................................................................................... 14
Running ALLPATHS – in brief ....................................................................................................................... 15
Example ................................................................................................................................................... 15
Pipeline errors ......................................................................................................................................... 16
The ALLPATHS assembly ............................................................................................................................. 16
Assembly as a graph ............................................................................................................................... 16
Graph features ........................................................................................................................................ 16
Repeats ............................................................................................................................................................... 16
Homopolymers ................................................................................................................................................... 17
SNPs and base errors .......................................................................................................................................... 17
ALLPATHS‐LG Manual 2
Flattening and Scaffolding ...................................................................................................................... 17
Assembly Results ..................................................................................................................................... 17
fasta format ........................................................................................................................................................ 17
efasta format ...................................................................................................................................................... 18
ALLPATHS Reference ......................................................................................................................... 19
ALLPATHS Cache for power users ............................................................................................................... 19
Creating the ALLPATHS Cache ................................................................................................................. 19
Importing fastq files of mixed PHRED encoding ...................................................................................... 20
Using the ALLPATHS Cache ..................................................................................................................... 21
ALLPATHS compilation options ................................................................................................................... 22
ALLPATHS pipeline – in detail ..................................................................................................................... 22
Key Features ............................................................................................................................................ 22
Directory structure – ALLPATHS_BASE .................................................................................................... 23
Targets .................................................................................................................................................... 23
Pseudo targets .................................................................................................................................................... 23
Target files .......................................................................................................................................................... 24
Evaluation mode ..................................................................................................................................... 24
Kmer size, K ............................................................................................................................................. 25
Parallelization ......................................................................................................................................... 25
Cross‐module parallelization .............................................................................................................................. 25
Parallelization of individual modules .................................................................................................................. 25
Logging .................................................................................................................................................... 25
References ............................................................................................................................................. 27
ALLPATHS‐LG Manual 3
Conventions
The following conventions are used in this manual.
Commands, filenames, directories and arguments are typeset in Courier.
Command‐line arguments are normally split one per line for clarity, listed below the actual command.
For example:
becomes
RunAllPathsLG
PRE=/assemblies
DATA=datadir
RUN=rundir
SUBDIR=attempt1
User‐supplied values are indicated by <description>. In the example below, the user should
provide a value for the target name.
TARGETS=<target name>
For example:
TARGETS=import
Introduction
ALLPATHS‐LG is a whole‐genome shotgun assembler that can generate high‐quality genome assemblies
using short reads (~100bp) such as those produced by the new generation of sequencers. The significant
difference between ALLPATHS and traditional assemblers such as Arachne is that ALLPATHS assemblies
are not necessarily linear, but instead are presented in the form of a graph. This graph representation
retains ambiguities, such as those arising from polymorphism, uncorrected read errors, and unresolved
repeats, thereby providing information that has been absent from previous genome assemblies.
Capabilities and limitations
ALLPATHS‐LG is a short‐read assembler. It has been designed to use reads produced by new sequencing
technology machines such as the Illumina Genome Analyzer. The version described here has been
optimized for, but not necessarily limited to, reads of length 100 bases.
ALLPATHS is not designed to assemble Sanger or 454 FLX reads, or a mix of these with short reads.
ALLPATHS‐LG Manual 4
ALLPATHS‐LG requires high sequence coverage of the genome in order to compensate for the shortness
of the reads. The precise coverage required depends on the length and quality of the paired reads, but
typically is of the order 100x or above. This is raw read coverage, before any error correction or filtering.
For small bacterial‐sized genomes, this translates to a fraction of an Illumina lane – the minimum the
machine is capable of without multiplexing. For larger genomes this translates into roughly one Illumina
HiSeq flowcell.
ALLPATHS‐LG requires a minimum of 2 paired‐end libraries – one short and one long. The short library
average separation size must be slightly less than twice the read size, such that the reads from a pair will
likely overlap – for example, for 100 base reads the insert size should be 180 bases. The distribution of
sizes should be as small as possible, with a standard deviation of less than 20%. The long library insert
size should be approximately 3000 bases long and can have a larger size distribution. Additional optional
longer insert libraries can be used to help disambiguate larger repeat structures and may be generated
at lower coverage.
The libraries must be ‘pure’, that is, they must consist of reads that do not contain any non‐genomic
portions from stuffers or similar constructions. Reads from jumping libraries may be chimeric, that is,
they may cross the junction point between the two ends of the insert that occurs in libraries produced
using the Illumina sheared library protocol.
Staying up to date with our blog
The best source of current news and information on ALLPATHS‐LG is our blog:
http://www.broadinstitute.org/software/allpaths‐lg/blog/
Here you will find announcements, an FAQ, links to the latest code, manual and test data, build
requirements and instructions, and information on how to get help from the developers of ALLPATHS‐LG
We recommend that our blog page should be your starting point whenever you have problems,
questions or are just looking for the latest version.
Requirements
To compile and run ALLPATHS‐LG you will need a Linux/UNIX system with at least 16 GB of RAM. We
suggest a minimum of 32 Gb for small genomes, and 512 Gb for mammalian sized genomes.
For the up to date list of requirements please see our General Build help here:
https://www.broadinstitute.org/science/programs/genome‐biology/computational‐rd/general‐
instructions‐building‐our‐software
You will need:
ALLPATHS‐LG Manual 5
The g++ compiler from GCC, version 4.7.0 or higher.
http://gcc.gnu.org/
The GMP library compiled with the C++ interface. Your GCC installation may already include GMP.
http://gmplib.org /
The Picard set of Java‐based command‐line utilities for SAM file manipulation available at
http://picard.sourceforge.net/
The graph command dot from the graphviz package.
http://www.graphviz.org/
Availability
The ALLPATHS source code is available for download via our blog at:
http://www.broadinstitute.org/software/allpaths‐lg/blog/
We do not issue official releases. Instead, please download the latest version from our nightly builds.
Only builds that pass our internal tests are made available in this way ‐ we do not release broken builds.
Getting Help
Please consult the FAQ available on our blog at:
http://www.broadinstitute.org/software/allpaths‐lg/blog/
Installation
For the up to date build instructions please see our General Build Instructions here:
https://www.broadinstitute.org/science/programs/genome‐biology/computational‐rd/general‐
instructions‐building‐our‐software
After you have downloaded the latest build, unpack it using tar. Then you can simply compile the
source code with configure and make. All of the source code will be in its own directory called
allpathslg-<revision>; we will refer to this as the AllPaths directory. For example, starting
from the root directory (the location of the downloaded file):
% cd allpathslg-<revision>.tar // move into the source directory
% make // build ALLPATHS‐LG
ALLPATHS‐LG Manual 6
% make install // install ALLPATHS‐LG
Troubleshooting
Of the above steps, the one most likely to fail is configure, which checks for the existence of various
commands and libraries in your environment. You may need to change your PATH or your
LD_LIBRARY_PATH. You may also need to run configure with flags. For a listing of all such
available flags, run configure --help.
Environment
After compilation, the executable binary files will be in the subdirectory bin of the allpathslg-
<revision> directory. You may want to add this directory to your PATH so that you can call the
ALLPATHS binaries from anywhere. Also modify your PATH to include the directories containing
addr2line and your chosen version of g++. You may need to change your LD_LIBRARY_PATH as
well.
ALLPATHS pipeline overview
ALLPATHS consists of a series of modules. Each module performs a step of the assembly process.
Different modules may be run, and in varying order, depending on the assembly parameters. A single
module called RunAllPathsLG controls the entire pipeline, deciding which modules to run and how
to run them. Although it is possible to run the individual modules manually, you should be able to
accomplish everything you need through RunAllPathsLG.
RunAllPathsLG module
RunAllPathsLG uses the Unix make utility to control the assembly pipeline. It does not call each
module itself, but instead creates a special makefile that does. Within RunAllPathsLG each
module is defined in terms of its source and target files, and the command line used to call it. A module
is only run if its target files don’t exist, or are out of date compared to its source files, or if the command
used to call the module has changed. In this way RunAllPathsLG can be run again and again, with
different parameters, and only those modules that need to be called will be. This is efficient and ensures
that all intermediate files are always correct, regardless of how many times RunAllPathsLG has been
called on a particular set of source data and how many times a module fails or aborts partway through.
ALLPATHS pipeline directory structure
The assembly pipeline uses the following directory structure to store its inputs, intermediates, and
outputs. The pipeline automatically creates the directories (if they don’t already exist) and populates
them. The names shown here are commonly used to refer to the directories, although command‐line
arguments determine the actual directory names.
REFERENCE/DATA/RUN/ASSEMBLIES/SUBDIR
ALLPATHS‐LG Manual 7
The meaning of each directory is given below. The data separation described is the ideal and
occasionally this is broken for convenience. Some files are duplicated between directories, but only in
the downward direction. All files within this directory structure are under the control of the pipeline.
The location of the pipeline directory structure is specified with the RunAllPathsLG command‐line
argument PRE.
Typically in the directory PRE there will be a number of REFERENCE directories, one for each organism
being assembled by ALLPATHS.
REFERENCE (organism) directory
The REFERENCE directory is so called because there should be one for each reference genome you use.
It is used to separate assembly projects by organism and possibly also by isolate (if, for example, you
want to use two different E.coli references) and is typically named after the organism. All assembly
projects for a given organism/isolate will be contained in that REFERENCE directory. All intermediate
files generated for use in evaluation that are independent of the particular assembly attempt will be
stored here and shared by all assemblies.
You do not need to supply a reference genome – ALLPATHS is, after all, a de novo assembler. But even
in de novo assemblies, the pipeline can perform useful evaluations at various stages of the assembly
process, so you should provide a reference genome if you have one (see “Import reference” below for
info on how to set up this file.) If you do not have a reference genome, simply create a single
REFERENCE directory for the organism you wish to assemble.
The REFERENCE directory may contain many DATA directories, each representing a particular set of
read data to assemble.
RunAllPathsLG argument: REFERENCE_NAME
DATA (project) directory
The DATA directory contains the original read data used in a particular assembly attempt. (This data is
stored in internal ALLPATHS formats: fastb, qualb, pairs.) It also contains intermediate files derived from
the original data that are independent of the particular assembly attempt – typically files used in
evaluation.
Each DATA directory may contain many RUN directories, each representing a particular attempt to
assemble the original data using a different set of parameters.
RunAllPathsLG argument: DATA_SUBDIR
RUN (assembly preprocessing) directory
The RUN directory contains all the non‐localized assembly files, that is, those intermediate files
generated from the original read data in preparation for the final assembly stage (LocalizeReadsLG
and beyond). It may also contain intermediate files used in evaluation that are dependent on the
assembly parameters chosen.
ALLPATHS‐LG Manual 8
RunAllPathsLG argument: RUN
ASSEMBLIES directory
The ASSEMBLIES directory contains the actual assembly (or assemblies). There is no argument for
naming this directory. It is actually named ASSEMBLIES.
SUBDIR (assembly) directory
The SUBDIR directory is where the localized assembly is generated, along with some assembly
intermediate and evaluation files.
RunAllPathsLG argument: SUBDIR
Required ALLPATHS arguments
The following command‐line arguments must be supplied:
PRE – the root directory in which the ALLPATHS pipeline directory will be created.
REFERENCE_NAME – the REFERENCE (organism) directory name ‐ described previously.
DATA_SUBDIR – the DATA (project) directory name ‐ described previously.
RUN – the RUN (assembly pre‐processing) directory name ‐ described previously.
SUBDIR – the SUBDIR (assembly) directory name ‐ described previously.
Preparing data for ALLPATHS
Before running ALLPATHS, you must prepare your data for import into the ALLPATHS pipeline. This task
will require you to gather the read data in the appropriate formats, and then add metadata to describe
them. If you are using a reference genome for evaluation, you will need that as well. This section
describes the required data formats.
Supported library constructions
Any input dataset should include at least one fragment library and one jumping library. A fragment
library is a library with a short insert separation, less than twice the read length, so that the reads may
overlap (e.g., 100bp Illumina reads taken from 180bp inserts.) A jumping library has a longer separation,
typically in the 3kbp‐10kbp range, and may include sheared or EcoP15I libraries or other jumping‐library
construction; ALLPATHS can handle read chimerism in jumping libraries. Note that fragment reads
should be long enough to ensure the overlap.
Additionally, ALLPATHS also supports long jumping libraries. A jumping library is considered to be long if
the insert size is larger than 20 kbp. These libraries are optional and used only to improve scaffolding in
mammalian‐sized genomes. Typically, long jump coverage of less than 1x is sufficient to significantly
improve scaffolding.
ALLPATHS‐LG Manual 9
ALLPATHS also accepts long unpaired reads (e.g., PacBio reads at 50x coverage), which are optional and
are used only to patch gaps in the later stages of the assembly process. Currently this is only tested for
small, bacterial‐sized genomes.
Any other type of library construction is not supported by ALLPATHS at this point.
Read orientation
Fragment library reads are expected to be oriented towards each other (inward):
Jumping library reads are expected to be oriented away from each other (outward), as a result of the
typical jumping library construction methods:
Long jumping library reads are expected to be oriented towards each other (inward), as a result of the
typical jumping library construction methods:
ALLPATHS input files
The DATA directory must initially hold files containing the sequenced reads, their quality scores and
information concerning their pairing. In addition a ploidy file must also be present. These files may
already exist if you are continuing or restarting an existing assembly, or may be assembled together
using tools provided with the ALLPATHS distribution.
Base, quality score, and pairing information files
The read libraries mentioned in the previous section must each be provided as a file containing the
bases, a file holding the quality scores, and a file with the pairing and library info. The specific file names
are:
<REF>/<DATA>/frag_reads_orig.fastb
<REF>/<DATA>/frag_reads_orig.qualb
<REF>/<DATA>/frag_reads_orig.pairs
<REF>/<DATA>/jump_reads_orig.fastb
<REF>/<DATA>/jump_reads_orig.qualb
<REF>/<DATA>/jump_reads_orig.pairs
The following long jump files are optional:
<REF>/<DATA>/long_jump_reads_orig.fastb
<REF>/<DATA>/long_jump_reads_orig.qualb
<REF>/<DATA>/long_jump_reads_orig.pairs
ALLPATHS‐LG Manual 10
As is the following long unpaired reads file:
<REF>/<DATA>/long_reads_orig.fastb
These files can be automatically generated from a set of BAM, fastq, or fasta files as described below.
The ploidy file
The file ploidy is a single‐line file containing a number. As the name suggests, this number indicates
the ploidy of the genome with 1 for haploid genomes and 2 for diploid genomes. Polyploid genomes are
not currently supported. The specific file name is:
<REF>/<DATA>/ploidy
Preparing ALLPATHS input files
The Perl script PrepareAllPathsInputs.pl can be used to automatically convert a set of BAM,
fasta, fastq, or fastb files to ALLPATHS input files. It will also optionally create the necessary ploidy
file. This is the easiest way to prepare data for ALLPATHS given a set of files from the Illumina platform.
The user must provide as input two comma‐separated‐values (.csv) files:
in_groups.csv
in_libs.csv
These describe, respectively, the locations and library information of the various files to be converted.
Accepted data file formats
Each data file must contain paired reads from a single library, but a library may be split over many files.
Typically a data file will represent a single lane of an Illumina flowcell.
As mentioned above, currently accepted formats are .bam, .fastq, and .fasta. The quality scores for
.fasta files are expected in corresponding .quala files.
For .fastq files you MUST check how the quality scores are encoded. By default it is assumed that the
quality scores are encoded using ASCII 33 to 126. If the quality scores are encoded using ASCII 64 to 126
you MUST specify the option PHRED_64=1 when running the conversion script (this is described
below). If you have fastq files with a mix of the two phred encodings refer to the end of the manual for
instructions.
in_groups.csv file
Each line in in_groups.csv provides, for each data file, the following information:
group_name: a UNIQUE nickname for this specific data set.
library_name: the library to which the data set belongs.
file_name: the absolute path to the data file. Wildcards ‘*’ and ‘?’ are accepted (but not in the
extension) when specifying multiple files as in the case of two paired or multiple unpaired fastq or fasta
files. Supported extensions are: ‘.bam’, ‘.fasta’, ‘.fa’, ‘.fastq’, ‘.fq’, ‘.fastq.gz’, and ‘.fq.gz’, all case‐
ALLPATHS‐LG Manual 11
insensitive. For ‘.fasta’ and ‘.fa’ it is expected that corresponding ‘.quala’ and ‘.qa’ files exist,
respectively.
Example in_groups.csv:
in_libs.csv file
Each line in in_libs.csv describes a library. The specific fields are:
library_name: matches the same field in in_groups.csv.
project_name: a string naming the project.
organism_name: the organism.
type: fragment, jumping, EcoP15, etc. This field is only informative.
paired: 0: Unpaired reads; 1: paired reads.
frag_size: average number of bases in the fragments (only defined for FRAGMENT libraries).
frag_stddev: estimated standard deviation of the fragments sizes (only defined for FRAGMENT
libraries).
insert_size: average number of bases in the inserts (only defined for JUMPING libraries; if larger
than 20 kb, the library is considered to be a LONG JUMPING library).
insert_stddev: estimated standard deviation of the inserts sizes (only defined for JUMPING
libraries).
read_orientation: inward or outward. Outward oriented reads will be reversed.
genomic_start: index of the FIRST genomic base in the reads. If non‐zero, all the bases before
genomic_start will be trimmed out.
genomic_end: index of the LAST genomic base in the reads. If non‐zero, all the bases after
genomic_end will be trimmed out.
ALLPATHS‐LG Manual 12
Here is an example in_libs.csv (NOTE: all the fields should be on a single line; that makes the lines
too long to show here, hence the '...'):
Running conversion script
Simplest example of a PrepareAllPathsInputs.pl run:
PrepareAllPathsInputs.pl
DATA_DIR=<full_path to REFERENCE DIR>/mydata
PICARD_TOOLS_DIR=/opt/picard/bin
where DATA_DIR is the location of the ALLPATHS DATA directory where the converted reads will be
placed, and PICARD_TOOLS_DIR is the path to the Picard tools needed for data conversion, if your
data is in BAM format. There are other options that can be specified:
IN_GROUPS_CSV – use a file other than ./in_groups.csv.
IN_LIBS_CSV – use a file other than ./in_libs.csv.
INCLUDE_NON_PF_READS – 1:(default) include non‐PF reads. 0: include only PF reads.
PHRED_64 – (for ‘fastq’ files only) 0:(default) – provided fastq’s have quality scores encoded with ASCII
33 to 126. 1: ASCII 64 to 126.
PLOIDY – generate the ploidy file. Valid values are 1 or 2.
HOSTS – list of hosts to use in parallel by forking (NOTE: forking to remote hosts requires password‐less
ssh access, e.g. using ssh‐agent/ssh‐add). Example: ’2,3.host2,4.host3’ which translates to:
• 2 processes forked on the localhost;
• 3 processes forked on host2;
• 4 processes forked on host3.
ALLPATHS‐LG Manual 13
The following options allow the user to select, randomly, a fraction of the total number of reads:
FRAG_FRAC – fraction of fragment reads to include, e.g. 30% or 0.3.
JUMP_FRAC – fraction of jumping reads to include, e.g. 20% or 0.2.
LONG_JUMP_FRAC – fraction of long jumping reads to include, e.g. 90% or 0.9.
GENOME_SIZE – estimated genome size for the purpose of coverage estimation.
FRAG_COVERAGE – fragment library desired coverage, e.g. 45. Requires GENOME_SIZE.
JUMP_COVERAGE – jumping library desired coverage, e.g. 45. Requires GENOME_SIZE.
LONG_JUMP_COVERAGE – jumping library desired coverage, e.g. 1. Requires GENOME_SIZE
(typically, only very low coverage is required for long jumps).
Note, however, that there are some restrictions on the above options. If you specify FRAG_FRAC or
JUMP_FRAC, you cannot also specify FRAG_COVERAGE or JUMP_COVERAGE. If you specify
FRAG_COVERAGE or JUMP_COVERAGE you must specify GENOME_SIZE, since both values are
necessary for the calculation of the read fraction to include.
After a successful run of PrepareAllPathsInputs.pl the necessary ALLPATHS input files should
be in place and ready for an assembly run to start.
Import reference
If you plan to perform evaluations, you can import a reference genome into the pipeline directory at the
same time as the read data. The reference genome to import is specified using the argument:
The reference genome must be supplied as two files: genome.fasta and genome.fastb. The fastb
file is a binary version of the fasta file. You can convert from fasta to fastb using the ALLPATHS module
Fasta2Fastb.
This argument is ignored if a reference genome already exists in the REFERENCE directory. It will not
cause an existing reference genome in the pipeline directory to be overwritten.
Once the reference has been imported into the REFERENCE directory, you can omit the
REFERENCE_DIR argument when running RunAllPathsLG.
Instead of using the REFERENCE_DIR argument, you may simply create the REFERENCE directory
and place the reference genome files in it. The reference genome files must be named:
genome.fasta and
genome.fastb
ALLPATHS‐LG Manual 14
Running ALLPATHS – in brief
Once the read data has been imported you may run the ALLPATHS pipeline as often as desired, each
time with different assembly parameters. Each time you run the ALLPATHS pipeline it will determine
which modules need to run (or re‐run) depending on the parameters you have chosen. Unless you want
to overwrite your previous assembly, specify a new RUN directory each time.
This section briefly describes the RunAllPathsLG arguments commonly used to run the ALLPATHS
pipeline. Complete descriptions of all arguments are provided in the ALLPATHS Reference.
evaluation mode ‐ Given a reference genome, the pipeline can perform evaluations at various
stages of the assembly process and of the assembly itself. To turn evaluation on, set
EVALUATION=STANDARD.
targets –The value of the TARGETS parameter determines the operations performed by the
pipeline:
TARGETS=full_eval Runs a version of the pipeline that includes additional evaluation
modules.
TARGETS=standard Runs a streamlined version of the pipeline that skips many of the
evaluation modules.
parallelization ‐ The pipeline has two levels of parallelization. It can run two or more modules
concurrently if their dependencies are independent. Many individual modules are also capable
of being parallelized via multithreading. By default, only multithreaded parallelization is on. See
the ALLPATHS Reference for more details.
Example
The TARGETS argument of RunAllPathsLG determines whether the ALLPATHS pipeline runs to
completion or imports the data and stops. To run an assembly using previously imported data use:
TARGETS=standard
For example, for data imported using PrepareAllPathsInputs.pl with DATA_SUBDIR=<user
pre>/staph/mydata use:
RunAllPathsLG
PRE=<user pre>
DATA_SUBDIR=mydata
RUN=myrun
REFERENCE_NAME=staph
TARGETS=standard
This will create (if it doesn’t already exist) the following pipeline directory structure:
<user pre>/staph/mydata/myrun
ALLPATHS‐LG Manual 15
Where staph is the REFERENCE directory, mydata is the DATA directory containing the imported
data, and myrun is the RUN directory.
Pipeline errors
The pipeline will stop when it encounters an error. There are two types of error that can occur:
rule consistency check error ‐ Before any modules are called, RunAllPathsLG checks to see
if it knows how to make all the output files for the given assembly parameters. If not, the
pipeline halts immediately before any modules are run, reporting the files that it does not know
how to make. Check and correct your arguments and try again.
runtime consistency check error ‐ After each module in the pipeline has completed, the pipeline
checks to see if correct output files were created. If any files are missing, the pipeline halts,
reporting the missing files and the module that failed to produce them. This most often occurs
when a module crashes. Check the log for an error message from the module in question.
Once the error has been identified and corrected, re‐run the RunAllPathsLG command. The pipeline
restarts at the point it previously failed.
The ALLPATHS assembly
Assembly as a graph
Unlike a conventional genome assembly, an ALLPATHS assembly is a graph. Edges in this graph represent
base sequences, and each path through the graph represents a possible solution to the assembly
problem. An ideal assembly would be a single edge, with occasional blips corresponding to SNPs in a
diploid genome. However, uncorrected sequencing errors, unresolved repeat structures, and assembly
algorithm inadequacies result in ambiguity. By representing the assembly as a graph we can capture this
ambiguity rather than arbitrarily choosing a solution and therefore losing information.
Graph features
A graph assembly consists of components and edges. A component is a collection of connected edges.
An assembly may consist of a number of components, scaffolded together as in a linear assembly.
In the following examples the edge lengths are not to scale. Purple represents long edges; red, medium
sized edges; black, short edges; and grey, very short edges.
Repeats
The graph below contains a 6.2 kb repeat that occurs 3 times in the genome. The repeat is longer than
the largest insert size available and so could not be resolved. However we do know the two possible
orderings of edges and can represent this in a graph.
ALLPATHS‐LG Manual 16
Homopolymers
With short reads, long homopolymer runs can be difficult to resolve. Rather than assuming a value for
the homopolymer length, they are represented as a loop of length 1 base.
SNPs and base errors
When the reads offer two seemingly equally possible alternatives for a base, we represent this as a small
bubble. This situation can arise from SNPs, in which case the bubble is “correct”, but it may also be due
to particularly hard‐to‐correct base substitution errors in the raw reads. In a conventional assembly,
bases of low quality would represent these ambiguities.
Flattening and Scaffolding
The graph assembly mentioned in the previous section provides the most complete description of the
genome. However, this graph incompatible with existing annotation and analysis tools and is also
troublesome to scaffold. To overcome these problems we attempt to flatten the graph, retaining as
much ambiguity as possible, then scaffold to create a (nearly) conventional assembly consisting of
scaffolded linear contigs.
The flattening and scaffolding processes are given elsewhere, but the end result is a pair of files – a
conventional fasta file plus a related efasta file. These files are described in detail in the next
section.
Assembly Results
The results of the assembly pipeline are given in the following two files in the assembly sub_dir
directory:
final.assembly.fasta
final.assembly.efasta
Both these files contain the final flattened and scaffolded assembly. The efasta, “enhanced” fasta,
file is a new format used by ALLPATHS and is based on the standard fasta file format.
fasta format
The fasta file contains contigs that have been scaffolded together, separated by n, where the number
of ns represents the best estimate of the gap size between contigs. A single n represents an unresolved
negative gap. Within each contig, an ambiguous base is represented by an N. For example, an A/T SNP
would become: ATGTCNTGTCG.
ALLPATHS‐LG Manual 17
efasta format
The efasta file contains contigs that have been scaffolded together, separated by N, where the
number of Ns represents the best estimate of the gap size between contigs. A single N represents an
unresolved negative gap. Within each contig, ambiguity is represented by an expression within a pair of
braces, {}. For example, an A/T SNP would become: ATGTC{A,T}TGTCG. An unresolved
homopolymer run of T, where the evidence suggested there should be 6, 7 or 8 Ts, would become:
GTCACTTTTTT{,T,TT}GCTGT. In this enhanced fasta format simple ambiguities that would
otherwise be lost are now retained. The efasta format can be easily flattened by picking the first
option for each ambiguity, resulting in the assembly given in the associated fasta file. Within the
braces the options are ordered in terms of decreasing likelihood.
ALLPATHS‐LG Manual 18
ALLPATHS Reference
ALLPATHS Cache for power users
The Perl script PrepareAllPathsInputs.pl that imports data to ALLPATHS is in fact a wrapper
around a few tools. It first creates a temporary cache of fastb/qualb files in <DATA>/read_cache/
for each data file described in in_groups.csv. Then, it automatically merges all these cached files
into each of the input files expected by ALLPATHS.
Alternatively, a non‐temporary cache can be created separately at a different location and it can work as
a repository of data for many different projects. The advantage of having a cache is that it separates the
time‐consuming step of converting data files (especially BAM files) to the fastb and qualb format from
the merging of the fastb and qualb files into ALLPATHS input files. This is useful, for example, when a
user wants to run different assemblies based on different subsets of the original data.
Creating the ALLPATHS Cache
The ALLPATHS cache stores all the information regarding the libraries and groups in two files in the
cache directory: libraries.csv and groups.csv. To build the cache the following commands
need to be run:
CacheLibs.pl
CACHE_DIR=<CACHE_DIR>
IN_LIBS_CSV=in_libs.csv
ACTION=Add
CacheGroups.pl
CACHE_DIR=<CACHE_DIR>
PICARD_TOOLS_DIR=/opt/picard/bin
IN_GROUPS_CSV=in_groups.csv
TMP_DIR=/large-tmp
HOSTS=’2,3.host2,4.host3’
ACTION=Add
The CacheLibs.pl command simply adds the library information in in_libs.csv to the cache
libraries.csv. The CacheGroups.pl command converts all the data files described in
in_groups.csv to fastb and qualb files in the cache, and adds the corresponding entries to the
cache groups.csv. The common options are:
CACHE_DIR – the full path to the cache directory. Can be omitted if the environment variable
ALLPATHS_CACHE_DIR is defined.
ACTION – Add, List, or Remove entries to, in, and from the cache.
ALLPATHS‐LG Manual 19
CacheLibs.pl options:
IN_LIBS_CSV – alternative file to the default ./in_libs.csv.
GroupLibs.pl options:
PICARD_TOOLS_DIR – the full path to the Picard tools needed for data conversion. Can be omitted
if the environment variable ALLPATHS_PICARD_TOOLS_DIR is defined.
IN_GROUPS_CSV – alternative file to the default ./in_groups.csv.
TMP_DIR – the full path of a local temporary directory; must be large if your data is large.
INCLUDE_NON_PF_READS – 1:(default) include non‐PF reads. 0: include only PF reads.
HOSTS – list of hosts to use in parallel by forking. Each fork converts a single data file (NOTE: forking
to remote hosts requires password‐less ssh access, e.g. using ssh‐agent/ssh‐add). Example:
’2,3.host2,4.host3’ which translates to:
• 2 processes forked on the localhost;
• 3 processes forked on host2;
• 4 processes forked on host3.
Finally, the contents of the cache can easily be listed by running:
CacheGroups.pl
CACHE_DIR=<CACHE_DIR>
ACTION=List
Importing fastq files of mixed PHRED encoding
When you have some fastq files encoded PHRED+64 and others PHRED+33 you will have to import them
to the cache separately.
First import the libraries with the CacheLibs.pl command as described in the previous section.
Then, create two files (in_groups_33.csv and in_groups_64.csv) with the corresponding
fastq groups.
Run the two following commands to import the groups to the cache:
CacheGroups.pl
CACHE_DIR=<CACHE_DIR>
IN_GROUPS_CSV=in_groups_33.csv
PHRED_64=False
…<plus other relevant options>
ALLPATHS‐LG Manual 20
CacheGroups.pl
CACHE_DIR=<CACHE_DIR>
IN_GROUPS_CSV=in_groups_64.csv
PHRED_64=True
…<plus other relevant options>
Now that you have successfully and correctly imported your groups to the cache, follow the instructions
in the next section to generate the ALLPATHS input files.
Using the ALLPATHS Cache
Once the cache is created it can be used to generate the ALLPATHS input files:
<DATA>/frag_reads_orig.fastb
<DATA>/frag_reads_orig.qualb
<DATA>/frag_reads_orig.pairs
<DATA>/jump_reads_orig.fastb
<DATA>/jump_reads_orig.qualb
<DATA>/jump_reads_orig.pairs
<DATA>/long_jump_reads_orig.fastb
<DATA>/long_jump_reads_orig.qualb
<DATA>/long_jump_reads_orig.pairs
<DATA>/long_reads_orig.fastb
The command to generate the ALLPATHS input files is:
CacheToAllPathsInputs.pl
CACHE_DIR=<CACHE_DIR>
GROUPS=”{12345AAXX.{1,2,3},67890ABXX.{6,7}}”
DATA_DIR=<DATA_DIR>
FRAG_FRAC=50%
JUMP_FRAC=34%
The options are:
CACHE_DIR – the path to the cache directory. Can be omitted if the environment variable
ALLPATHS_CACHE_DIR is defined.
DATA_DIR – the full path to the ALLPATHS DATA directory where the input files will be placed.
GROUPS – a list of the groups to include as inputs.
IN_GROUPS_CSV – file including the groups description. Optional alternative to GROUPS.
ALLPATHS‐LG Manual 21
FRAG_FRAC – fraction of fragment reads to include, e.g. 30% or 0.3.
JUMP_FRAC – fraction of jumping reads to include, e.g. 20% or 0.2.
LONG_JUMP_FRAC – fraction of long jumping reads to include, e.g. 90% or 0.9.
FRACTIONS – (use with GROUPS only) list of fractions, one per group, e.g. ”{0.5,30%,100%}”.
GENOME_SIZE – estimated genome size for the purpose of coverage estimation.
FRAG_COVERAGE – (requires GENOME_SIZE) fragment library desired coverage, e.g. 45.
JUMP_COVERAGE – (requires GENOME_SIZE) jumping library desired coverage, e.g. 45.
LONG_JUMP_COVERAGE – (requires GENOME_SIZE) jumping library desired coverage, e.g. 1.
(typically, only very low coverage is required for long jumps).
COVERAGES – (use with GROUPS only, requires GENOME_SIZE) list of coverages, one per group,
e.g. ”{45,50,2}”.
LONG_READ_MIN_LEN – (default 500) this sets the threshold for what qualifies as a long unpaired
read (e.g. PacBio reads).
As in PrepareAllPathsInputs.pl, there are some restrictions on the above options. If you
specify FRAG_FRAC, JUMP_FRAC, or FRACTIONS, you cannot also specify FRAG_COVERAGE or
JUMP_COVERAGE, or COVERAGES. If you specify FRAG_COVERAGE, JUMP_COVERAGE, or
COVERAGES you must specify GENOME_SIZE, since both values are necessary for the calculation of
the read fraction to include. If you specify one of FRACTIONS and COVERAGES lists you must specify
a GROUPS list and supply one fraction or coverage entry for each group.
After a successful run of CacheToAllPathsInputs.pl the necessary ALLPATHS input files should
be in place and ready for an assembly run to start.
ALLPATHS compilation options
The following command‐line options may be appended to make when building ALLPATHS:
-j<n> Split the compilation into n parallel processes. If you set n equal to the number of CPUs on
your machine, it will speed up compilation approximately n‐fold. See Installation for an
example.
ALLPATHS pipeline – in detail
Key Features
The ALLPATHS pipeline incorporates the following key features:
ALLPATHS‐LG Manual 22
Runs only those modules that are required for a particular set of parameters.
Ensures intermediate files are always consistent.
If the parameters for a module change, reruns only the changed module and modules that
depend on its output.
In the event of a problem, restarts at the point the problem occurred.
Supports easy parallelization by allowing modules that don’t depend on each other’s output to
run concurrently.
Can easily be run up to any point.
Can initially exclude modules that are not required for the assembly process (evaluation
modules for example), then easily run them once the assembly is complete.
Determines if it has all the necessary input files and knows how to build all the requested output
files before starting any modules. Stops immediately if there is a problem.
Directory structure – ALLPATHS_BASE
In addition to using the command‐line argument PRE to specify the location of the pipeline directory,
you may optionally also use ALLPATHS_BASE. The pipeline directory location is either:
PRE
or
PRE/ALLPATHS_BASE
Targets
The pipeline determines which output files it needs to generate by means of a list of targets. If a
particular target file is requested, then the modules required to create both it, and any intermediate
files it depends on, will be run in the correct order. Only these modules will be run. Further, if any
required intermediate files already exist and are up to date with respect to the files that they in turn
depend on, then the call to the module required to build them is skipped. This holds true for the final
target file or files – if they already exist and are up to date then nothing will be done.
You can specify the target files to build in two ways. The simplest is to use one of the predefined pseudo
targets that represent a set of useful target files – much like pseudo targets in Make. The second is to
specify a list of individual files that the pipeline knows how to make. Both methods may be used at the
same time.
If you ask for a target file that the pipeline doesn’t know how to make you will get an error message.
Pseudo targets
This is the best way to control which files the pipeline will create. The pseudo target value is passed to
RunAllPathsLG using:
There are 3 possible pseudo targets:
none – no pseudo targets, only make explicitly listed target files (see below).
ALLPATHS‐LG Manual 23
standard – create the assembly and selected evaluation files.
full_eval – create the assembly and additional evaluation files.
The default target is standard.
Target files
Individual files may be specified as targets instead of, or in addition to, the pseudo targets. Lists of target
files in each pipeline subdirectory are passed to RunAllPathsLG using:
Multiple target files may be passed in the following manner:
TARGETS_RUN=”{target1,target2,target3}”
The list of valid target files changes based on the assembly parameters chosen.
Evaluation mode
Given a reference genome, the pipeline can perform evaluations at various stages of the assembly
process.
Certain evaluations have the potential to alter the assembly, as they require reference genome data to
be incorporated into data structures used by the assembly process. Any such perturbation of the
assembly should be neutral but will have a stochastic effect on the result. Such ‘unsafe’ evaluations
allow much more detailed information to be gathered about the assembly process and are extremely
useful during development, but can be considered “cheating” from the point of view of de novo
assembly.
The evaluation mode used is controlled by:
EVALUATION=<evaluation mode>
There are three evaluation modes:
NONE – do not evaluate/no reference is available.
BASIC – basic evaluation that does not require a reference.
STANDARD – run evaluation modules using a supplied reference.
FULL – turn on in‐place evaluation in certain assembly modules. Does not perturb assembly.
ALLPATHS‐LG Manual 24
CHEAT – run in‐place evaluations that potentially perturb the assembly (in a neutral fashion),
but allow a more detailed analysis.
The default mode is BASIC.
Kmer size, K
The user should not adjust the kmer size from the default value of K=96.
The relationship between kmer size K and read size is not a direct one in ALLPATHS‐LG, unlike in may
other assemblers. ALLPATHS‐LG actually uses a number of different sizes of K internally, and because of
this, it is not intended that users change the K values for an assembly.
Parallelization
Given sufficient memory, it is possible to parallelize the pipeline in order to reduce runtime. Two forms
of parallelization are possible and both may be used at the same time.
Crossmodule parallelization
Modules in the pipeline that do not depend on each other may be run concurrently. This functionality is
provided by make, which is used by RunAllPathsLG to execute the pipeline. It is equivalent to using
the option –j<n> when compiling the ALLPATHS source code. No checks are made to ensure that there
is enough memory to run multiple ALLPATHS modules at the same time. Set the maximum number of
modules that can run concurrently using:
MAXPAR=<n>
The majority of the pipeline now uses parallel threading, so in most cases there is little to be gained in
setting this value about 1.
Parallelization of individual modules
Many of ALLPATHS’s modules have been engineered to run with parallel threading. This form of
parallelization is independent of the module parallelization described above. The level of parallelization
can be controlled using the argument to RunAllPathsLG:
THREADS=<n>
For maximum performance, set this value to the number of processors available – but be wary of
exceeding available memory as the number of threads increases. Due to hardware restraints (such as
I/O limiting and heap contention) you will find diminishing returns in runtime improvement as the
number of threads increases.
By default the pipeline will attempt to use all available processors.
Logging
In addition to standard out, the output from each ALLPATHS module is captured to file. In each pipeline
directory there exists a subdirectory named makeinfo that contains various logging files plus
ALLPATHS‐LG Manual 25
metadata used by the pipeline to control and track progress. Every single file produced by the pipeline
will have two log files associated with it. For example, the file hyper.fasta will have the following log
files in SUBDIR/makeinfo:
hyper.fasta.cmd
hyper.fasta.DumpHyper.out
The .cmd file contains the command used to generate hyper.fasta. The .out file contains the
captured output of the module used to create hyper.fasta. In this case the module is called
DumpHyper, as you would see from looking at the file hyper.fasta.cmd.
ALLPATHS‐LG Manual 26
References
Gnerre S, MacCallum I, Przybylski D, Ribeiro F, Burton J, Walker B, Sharpe T, Hall G, Shea T, Sykes S,
Berlin A, Aird D, Costello M, Daza R, Williams L, Nicol R, Gnirke A, Nusbaum C, Lander ES, Jaffe DB. High‐
quality draft assemblies of mammalian genomes from massively parallel sequence data Proceedings of
the National Academy of Sciences January 2011 vol. 108 no. 4 1513‐1518
MacCallum I, Przybylski D, Gnerre S, Burton J, Shlyakhter I, Gnirke A, Malek J, McKernan K, Ranade S,
Shea TP, Williams L, Young S, Nusbaum C, Jaffe DB. ALLPATHS 2: small genomes assembled accurately
and with high continuity from short paired reads. Genome Biology 2009, 10(10):R103.
Butler J, MacCallum I, Kleber M, Shlyakhter IA, Belmonte MK, Lander ES, Nusbaum C, Jaffe DB.
ALLPATHS: De novo assembly of whole‐genome shotgun microreads, Genome Res. May 2008 18:810‐
820.
ALLPATHS‐LG Manual 27