variant calling on amplicon data

Question

[Richard]

Hi all,

Does anyone have a recommended approach for variant calling (SNVs and indels) on human amplicon illumina reads? My bams are aligned with BWA and have had the chastity failed and duplicate reads marked.

If possible, I'd like to be able to apply a variant caller that can be instructed to ignore the duplicate status of the reads, that way we can keep the original bam that is created by our pipeline as well as keep the duplicate reads marked in case that information is needed later.

Thanks!

[Zaag]

Amplicons are duplicates by definition. Why mark them in this scenario?

[Richard]

For now, it is a limitation of our pipeline - that all bams are duplicate marked when they come through.

[User 59]

Then fix the pipeline?

[Brice Sarver]

Why won't a standard variant calling approach work for your data? If you have BAMs, kick them over to GATK. Reads that fail a duplicate filter (such as the one introduced using the mapping-based approach in Picard) will be handled appropriately, as per GATK's best practices pipeline.

Answer 1

[Zaag]

I just align them with BWA, sort with samtools and call variants with GATKs Haplotypecaller.