Execution

How to Run cgpCaVEManWrapper

1. Single Host Execution
2. Farm style execution

Option 2 provides more control over individual steps in the process. For CaVEMan option 2 is preferred as option 1 results in much longer runtimes.

Single host execution

Single host execution does require a reasonable level of resources. Minimum specification (Based on WGS at 30x coverage):

• CPU 32
• RAM (Gb) 4 / CPU (Some data may require far higher resources)

NB The code will recover from failure and restart at the last successful section.

Farm style execution

For users with access to a compute farm it is possible to break down the execution into the component parts (see Overview).

There are 9 steps:

1. setup

• Generate configuration files for the remaining steps.
• Command suffixed with -process setup -index 1

2. split

• Split the genome into chunks by readsize and hard stop forced by contig ends
• Command suffixed with -process setup -index X
• X is between 1 and the number of contigs in reference.fa.fai

3. split_concat

• Concatenate the split file sections into a single split section reference file
• Command suffixed with -process split_concat -index 1
• Generates splitList

4. mstep

• One job per line in splitList, builds a profile of the section of the genome using various covariates
• Command suffixed with -process mstep -index N
• N is between 1 and the number of lines in splitList

5. merge

• Merge the mstep profiles into a single profile for the whole genome
• Command suffixed with -process merge -index 1
• Produces probs_arrand covs_arr

6. estep

• One job per line in splitList, uses the profile built in the mstep, combined with sequence data and copy number to assign a probability to each possible genotype at each position.
• Command suffixed with -process estep -index N
• N is between 1 and the number of lines in splitList
• Positions where germline mutation probabilities total more than the -snp-cutoff are output to <split_section>.snps.vcf
• Positions where somatic mutation probabilities total more than the -mut-cutoff are output to <split_section>.muts.vcf

7. merge_results

• Merges all output results files into per-genome files names T_vs_N.*.vcf.
• Command suffixed with -process merge_results -index 1

8. add_ids

• Appends UUIDs to each entry in both the snps and the muts output VCF files.
• Command suffixed with -process add_ids -index 1

9. flag

• Applies filters to the T_vs_N.muts.vcf.gz file. Produces T_vs_N.muts.flagged.vcf.gz
• Command suffixed with -process flag -index 1

Flagging is described in further detail on the cgpCaVEManPostProcessing wiki

Overview of cgpCaVEManWrapper workflow