- Summary
- Rerun SVCandidateGeneration without recreating the SVLocus graph
- Comparing VCF output between runs
- Options to accelerate a small test case
This page describes debug/analysis capabilities which are especially useful to full Manta runs -- for instance for scenarios when iterating on a general improvement to the methods. When debugging a single SV, see the related page on Debugging a single SV in Manta
This is useful if you're working on a component of candidate generation/scoring which doesn't impact graph creation, and frequently rerunning a test. To use this option provide the '–rescore" option to runWorkflow.py. When this is provided candidate generation and scoring will always be re-run, but the graph will only be created if it doesn't already exist. Example
${RUN_DIR}/runWorkflow.py -j 24 --rescore
To assist in evaluating the quality of predictions from a full run compared to a stable benchmark (master, etc), there is a manta utility script to suppress some of the noise expected from a simple 'diff' of two manta vcfs. It takes two vcf files as arguments. These can be gzipped or uncompressed. A usage example is:
${MANTA_GIT_CLONE_DIR}/scratch/util/compareMantaVcfs.bash ../m67_test/m67_12_redo_control/results/variants/diploidSV.vcf.gz m63/results/variants/diploidSV.vcf.gz | grep -v MaxDepth | less
If not running an analysis on single small genome segment (see Debugging a single SV in Manta) there are various options to make small bam subsegments run a bit faster:
At config time you can reduce/increase the total number of tasks by making each job do more/less :
--scanSizeMb=scanSizeMb
Maximum sequence region size (in Mb) scanned by each
task during SV locus graph generation. (default: 12)
--candidateBins=candidateBins
Provide the total number of tasks which candidate
generation will be sub-divided into. (default: 256)
At run time you can shut down stderr logging, this log is replicated to $runDir/workflow/pyflow.data/logs/pyflow_log.txt so there is no loss of information. To do so, provide 'runWorkflow.py' with the "--quiet" option.