Pindel lower the mutation rate in indel calling

Question

[jinxinhao1988]

Dear all,

I have a question with calling indels with Pindel. First I use BWA mem in mapping the reads with hg19 reference with default parameter. Then I use pindel to call indels, since I have the standard answers of my data. The mutation rate of indels are all below the answers. I use 1% and 500X depth as a cut-off value. Therefore I missed some high confidence somatic indels (should be above 1% in mutation rate).

For now, we think that it may be the parameter we set in BWA mapping. Maybe we should change the mismatch settings to a lower one to allow more mismatch. Does anyone have encountered this situation and could give us some suggestions? It would be highly appreciated. Thank you for your time.

Answer 1

[Kevin Blighe]

I think that you will have better luck by configuring the settings of pindel. For example, after you run pindel, convert the output to VCF and configure parameters like --min_coverage, --het_cutoff, and --hom_cutoff:

pindel2vcf --pindel_output_root pindelOutput/ -r hg19.fasta -R GRCh37 -d 12121990 --min_coverage 1 --het_cutoff 0.1 --hom_cutoff 0.9 --vcf out.vcf

Also look at the -E/--sensitivity parameter that is passed to the pindel command itself. Look at all other options, too.

Kevin

[jinxinhao1988]

Thank you Kevin for your reply. The main problem is that the pindel wrongly estimated the VAF of indels, so I think that the filteration may not help much. I have tried to adjust the -E parameter, It didn't change much. I have contacted with the author of the software, and he replied that the pindel may cause this problem because of its stringent calculation of ALT reads. I have turned to use Mutect2 to call somatic mutation, and I'm satisfied with the results. But still the GATK running time is SO SO long.

[Kevin Blighe]

Yes, great that you contacted the author. Also, yes, GATK can be very slow! GATK4 should have CPU parallel processing capability, though.