The SRAssembler repository is the distribution point of our code for selective local assembly of genomic regions matching a homologous query protein or DNA sequence.
SRAssembler (Selective Recursive local Assembler) is a modular pipeline program that can assemble genomic DNA reads into contigs that are homologous to a query DNA or protein sequence. SRAssembler first collects the reads that can be locally aligned to the query sequence and assembles them into contigs. Additional reads are then found by aligning against the contigs assembled in the previous round. This in silico chromosome walking strategy recursively gathers reads that are associated with regions of interest. The gene structures of the final contigs are predicted by spliced alignment and ab initio gene finding programs.
Hopefully this code will be useful to you if you
Detailed instructions for installation from source are available in the INSTALL document. The simplest way to get going is to use the SRAssembler Singularity container, e.g. as follows:
cd
git clone https://github.com/BrendelGroup/SRAssembler
cd SRAssembler/demo
wget https://BrendelGroup.org/SingularityHub/sra.sif
alias rws="singularity exec -e -B ~/SRAssembler/demo ~/SRAssembler/demo/sra.sif"
rws ./xtest defaultpath
In the above example, you clone this repository into your Linux home directory, go into the SRAssembler/demo directory that has been created, download the SRAssembler Singularity container, define the bash alias rws ("run with singularity"), and check that everything works by running the xtest command with the defaultpath option (which will use the binaries from within the container).
Of course this assumes that you have Apptainer/Singularity installed on your system. Check whether there is package built for your system. Otherwise, follow the instructions to install Singularity from source code.
SRAssembler is run at the command line with path/to/SRAssembler [options]
. It can also take advantage of multiple processors on machines with OpenMPI by using mpirun -np $number_of_processors path/to/SRAssembler_MPI [options]
.
If you are using the Singularity image instead of compiling the SRAssembler binaries, you can run the SRAssembler program with singularity run --cleanenv --bind $(pwd) path/to/sra.sif [options]
. To use the Singularity image with multiple processors, run singularity exec --cleanenv --bind $(pwd) path/to/sra.sif mpirun -np $number_of_processors SRAssembler_MPI [options]
.
SRAssembler is a workflow program that controls and directs the activity of other programs (e.g., a read aligner, an assembler, a spliced aligner, an ab initio gene finder). SRAssembler requires both a parameter configuration file and command-line arguments. The parameter file contains settings for the programs under SRAssembler's control, while the command-line options have broader effects on SRAssembler's behavior.
-q query_file : FASTA file containing the DNA or protein query sequence. Required unless only pre-processing reads (-P).
-t query_type : Query sequence type. Accepts "protein" or "dna" (default: "protein").
-p parameter_file : Configuration file for the programs used by SRAssembler. SRAssembler allows user to specify the parameters of the programs it controls. The parameter configuration file has sections for each program, marked by a header line in square brackets. In some cases, such as the aligner Vmatch, there are multiple headers for a program, distinguishing different parameters for different usages. The example parameter file contains all of the program sections that SRAssembler recognizes, even though not all of them will be used in any one run of SRAssembler (e.g., GeneSeqer and GenomeThreader are mutually exclusive). The example file also contains comment lines marked with '#' that explain what each section is for. Copying and modifying this example file for your own runs is encouraged. A parameter file is required unless only pre-processing reads (-P flag).
-o output_directory : Output directory (default: "./SRAssembler_output").
-T temp_directory : Directory for temporary files. SRAssembler writes and reads many small files that are removed at the end of a run. Default is "/tmp", but you may get speed enhancements by using "/dev/shm" to keep these files entirely in memory.
-l library_file : Library definition file for sequencing reads. Required if the -1 option is not used. Paired-end reads must be in two sorted files rather than one interleaved file. Each library is specified in its own section, headed by a "[LIBRARY]" line. See the demo directory for example files. Each library includes the following items, although default values will be used for some items if they are absent:
-1 left_end_file : The single-end reads file, or the left-end reads file for paired-end reads. Required if -l option is not used.
-2 right_end_file : The right-end reads file for paired-end reads.
-Z insert_size : The insert size of the paired-end reads specified by -1 and -2, required if -2 option is used.
-r preprocessed_directory : Directory in which to store or from which to retrieve processed reads. Before the recursive search and assembly of SRAssembler, the input read files specified for each Library are split into multiple files and indexed by the alignment program, Vmatch. This pre-processing step only needs to happen once. When SRAssembler is run multiple times, pre-processed reads in this directory that correspond to the input libraries will be used again without needing to be processed. If not specified, defaults to "output_directory/processed_reads".
-R num_reads_per_split : Number of reads in each split file. While using larger split file parts tends to be faster, there should be enough parts to take advantage of all the processors available if you are using MPI parallelization. Note that if pre-processed reads already exist for a library, they will not be re-processed into new sizes due to this option. Defaults to "500000".
-P : Run the read processing step only.
-A assembler_program : The read assembler program used by SRAssembler. Use "0" for SOAPdenovo2 or "1" for ABySS, defaults to "0".
-k kmer_sizes : The k-mer sizes used by the assembler. SRAssembler will test multiple k-mer sizes and select the best for each round based on spliced-alignment results. The format for specifying which k-mers to test is start_k:interval:end_k. The start_k and end_k values must be odd, and the interval value must be even. For example, the default of "15:10:45" means k-mer values 15, 25, 35, and 45 will be tested.
-S spliced_alignment_program : The spliced alignment program used by SRAssembler. Use "0" for GeneSeqer or "1" for GenomeThreader, defaults to "1".
-s species : Species model for spliced alignment. Options are "human", "arabidopsis", "mouse", "rat", "chicken", "drosophila", "nematode", "fission_yeast", "aspergillus", "maize", "rice", and "medicago". Defaults to "arabidopsis".
-G gene_finding_program : The ab initio gene finding program used by SRAssembler. Use "0" to not use a gene finder or "1" for SNAP, defaults to "0".
-i query_contig_minimum : Minimum length of contigs used for finding more reads during chromosome walking. Increasing this may make SRAssembler searches slower, but more stringent. Defaults to "200".
-m minimum_contig_length : Minimum length of contig to accept as a potential hit for the query. Defaults to "200".
-M maximum_contig_length : Maximum length of contig to assemble before discontinuing extension of that contig. If an assembled contig is larger than the maximum_contig_length, it is trimmed at both ends to equal the maximum_contig_length and copied to the candidate-long-contig file. The candidate-long-contigs are aligned to the assembled contigs of the following round, and if they match (confirming a correct assembly) they are moved to the permanent long-contigs file and removed from the query file for the next round. The current set of matched reads are aligned to the long-contigs, and matching reads are removed (not used for assembly in future rounds). Defaults to "10000".
-e minimum_score : Minimum spliced alignment score to accept as a potential hit for the query. Defaults to "0.5".
-c minimum_coverage : Minimum fraction of coverage of the query by the spliced alignment. Defaults to "0.8".
-n num_rounds : Maximum number of rounds of chromosome walking. Defaults to "10".
-E extra_rounds : Number of additional rounds of recursion to perform once a hit contig is found. This option may be useful when automating many SRAssembler runs in which extending the contig beyond the bounds of the query (e.g., into the UTRs surrounding a CDS) is desirable, but the number of rounds necessary is unpredictable. This does not always lead to longer hit contigs, and may in fact result in a worse assembly. This option will not cause SRAssembler to exceed the number of rounds set by -n. Defaults to "0".
-a assemble_round : The round in which to start read assembly, defaults to "1". If read coverage depth is good, starting in round 1 is fine. In some cases, however, using the reads found in round 1 as queries to find additional reads before doing contig assembly can lead to better results.
-b clean_frequency : The frequency with which to remove unrelated contigs and reads from the recursive search. SRAssembler will periodically remove assembled contigs that do not align to the query sequence, and remove reads that are not being assembled into contigs. More frequent cleaning can make SRAssembler run faster, but may aggressively remove reads that would be relevant to correct assembly of a homolog of the query. The default of "4" means that SRAssembler will purge contigs and reads after three rounds of not doing so. This is not the same as doing so on every round evenly divisible by four, because the count from the last cleaning round can be reset by a forced cleaning when too many contigs are assembled, as explained under -d, or when continuing a run from pre-existing results.
-d contig_limit : The minimum number of assembled contigs to automatically trigger removal of unrelated contigs and reads. This can prevent slowdown or crashing in cases where an excessive number of reads are found due to things like common domains or transposable elements. If set to "0", do not remove unrelated contigs and reads except as scheduled by -b. Defaults to "500".
-j taboo_file : FASTA file containing DNA or protein sequences used to taboo reads and prevent them from being used for assembly. SRAssembler may struggle with reads from genomes with lots of transposons or other repetitive elements, or with assembling homologs to a query that contains a very common domain. With this option, SRAssembler will start by searching for reads that match the taboo_file, and exclude them from all assemblies.
-J taboo_type : Taboo file type. Accepts "protein" or "dna" (default: "protein").
-z masking_round : The round in which to start masking low-complexity regions of intermediate contigs before searching for reads. Reads with low-complexity regions that are assembled into contigs can wreak havoc on SRAssembler function. The number of irrelevant reads found in the next round due to shared low-complexity sequences can dwarf the number of relevant reads, and can slow down or crash the assembler. By default this option is set to "1", and SRAssembler uses NCBI's DustMasker to mask low-complexity regions in the intermediate contigs in every round. If you want to disable this feature, set this option to "0".
-x tip_search_round : The round in which to start masking the center of query contigs. If you find SRAssembler runs slowly due to finding excessive numbers of new reads, it can be helpful to mask the center of the intermediate contigs, leaving only the contig 'tips' to match against new reads. This causes SRAssembler to prioritize finding reads that can extend the contigs, at the cost of possibly missing some reads that might contribute to a hit contig. When set to "0", no masking of the center of contigs will occur. Defaults to "2", a value that ensures that contigs derived from the inital round of searching using the query sequence to find reads (rather than intermediate contigs) are used fully before masking. If using center masking, actively experiment with this option, as it can have a significant effect on whether optimal hit contigs are found.
-X tip_search_length : Length of intermediate contigs to leave unmasked at both ends as queries to find new reads. When using this option, it should at minimum be as long as the match length requirements set in the config file input with -p. When set to the default of "0", no masking of the center of contigs will occur.
-f : Do not check for contigs that meet the completion thresholds. Force SRAssembler to continue chromosome walking for the number of rounds specified with -n. This is useful if you want to build contigs longer than just the coding region in order to capture promoter regions or nearby elements.
-y : Do not resume chromosome walking from existing SRAssembler results in the output directory specified by -o. By default, SRAssembler will continue the previous run if the -n option is larger than the final round in the output directory. This behavior is useful if you want to extend the hit contigs for only a few rounds or change your completion thresholds to be more selective. Using the -y option will cause SRAssembler to overwrite the results in the output directory and start from round 1.
-v : Verbose mode. Not only is output to stdout more verbose, but several files in the aux/ directory that are normally removed to keep things tidy are left behind. This mode is typically only used for debugging.
-h : Print an SRAssembler usage synopsis and exit.
If no output directory was specified with -o, SRAssembler will create a directory called "SRAssembler_output" within the working directory. If no processed reads directory was specified with -r, the processed reads are created in a directory called "processed_reads" within the output directory. If no temp directory was specified with -T, SRAssembler will use "/tmp". Assuming these defaults, SRAssembler will create the following directories and files:
The demo/ directory contains the necessary files to test the functionality of most of SRAssembler's options. The bash script xtest runs SRAssembler several times using various command-line options and compares the results to the summary files in demo/standards/. This confirms the installation and functionality of all the SRAssembler dependencies. The bash script xquicktest runs just tests 1 and 4, confirming the basic functionality of the SRAssembler and SRAssembler_MPI executables. The demo/ directory also contains two library configuration files and a sample parameter configuration file. Examining these files is a good way to familiarize yourself with setting up SRAssembler runs.
Within demo/input/ are two FASTA query files and four FASTQ reads files. The query files are the transcript sequence of At1g01950.1, an Arabidopsis thaliana gene located at Chr1:325316-330619, and the amino acid sequence of the homologous rice protein LOC_Os06g04560.1. We used SAMTools' wgsim program to simulate two libraries from region 300,000..400,000 of chromosome 1 of Arabidopsis, with insert sizes 200bp and 1kb. Each library has 50,000 paired reads. The base error rate is 0.002 and the fraction of indels is 0.001.
Running the tests is as simple as executing bash xtest
. Each test will output progress to standard out, and after each test the script will report whether the test's summary file matches the one in the standards/ directory. At the end of the script, the comparison of all the tests' results to the standards will again be reported for easy viewing. To test the functionality of the Singularity image, run singularity exec -e -B $(pwd) path/to/SRAssembler.sif ./xtest defaultpath
from within the demo/ directory. This will bind the demo directory to the Singularity instance, allowing it to read and write, but cause it to use only its internal SRAssembler software to execute the tests.
The SRAssembler package conforms to our RAMOSE philosophy and promise: we made every attempt to make the software such that you can produce reproducible, accurate, and meaningful results; our software is open (source), designed to be scalable, and easy to use (so that a typical genome laboratory should be able to run it). And, of course, our software should be ramose ("having many branches") - it's yours to use, modify, improve, and enhance under the GNU General Public License v3.0.
Thomas W. McCarthy, Hsien-Chao Chou, and Volker P. Brendel (2019) SRAssembler: Selective Recursive local Assembly of homologous genomic regions. BMC Bioinformatics 20, 371.
Please direct all comments and suggestions to Volker Brendel at Indiana University.