stefinfection / rufus Goto Github PK

Where are the genentech ones?

Re-read the paper to see if any other two letter codes that we need to incorporate in our analyses?

Where are the low and medium quality vcf files?

Steps to reproduce variants not called by RUFUS that are in the gold-standard data set:
(Notes from S. Gardiner)

Files to reproduce:

So, I've been looking at the RUFUS run from the merged bams of EA/NC/LL with this file path:
/scratch/ucgd/lustre-work/marth/u0880188/smaht/hcc1395_seqc2/merged_runs/EA_NC_LL

How to identify variants that do have contigs made, but don't have calls in vcf:

But, the way I did it previously was I ran bedtools coverage on this file that contains contigs:
/scratch/ucgd/lustre-work/marth/u0880188/smaht/hcc1395_seqc2/merged_runs/EA_NC_LL/EA_NC_LL_3_merged_tumor.bam.generator.V2.overlap.hashcount.fastq.bam at the specific sites from the validated vcf that RUFUS failed to call.
This gave me this file: /scratch/ucgd/lustre-work/marth/u0880188/smaht/hcc1395_seqc2/merged_runs/EA_NC_LL/contig_depth.txt
Where i then just used a python script to pull out locations that had at least a coverage of 1.
Here is a tsv file of that:
/scratch/ucgd/lustre-work/marth/u0880188/smaht/hcc1395_seqc2/merged_runs/EA_NC_LL/contig_variants_validated.tsv

Looks like there were 1209 variants

Evaluate existing performance of RUFUS:

Option A

1. Find out sample IDs for PCAWG data used
2. Get bams and vcfs corresponding to those samples
3. Set up experimental design of N-T pairs
4. Run RUFUS on pairs
5. Accumulate calls and quantify sensitivity & non-overlapping calls (true positives?)
6. Generate figures to compare first round of PCAWG calls vs. my second round