Unexpected Discrepancy in iSNV Detection Between Workflows With and Without Kraken2 Filtering

[YangJingqii]

Hi,

I hope this finds you well. I am writing to seek your guidance regarding an unexpected observation in our RSV analysis workflows.

I have implemented two analysis pipelines for processing RSV samples:

A basic workflow: fastp trimming → bwa mapping → lofreq variant calling
An extended workflow: identical to the first, but with an additional kraken2 filtering step (using Pneumoviridae[11244]) before bwa.
I've noticed an intriguing discrepancy between these workflows. The extended workflow with kraken2 filtering identified 1542 iSNVs, while the basic workflow only detected 229 iSNVs. Interestingly, when comparing the depth coverage in the BAM files from both workflows, they show nearly identical patterns except for the last 26 positions. Further investigation using IGV revealed that the additional variants found in the kraken2 workflow are actually present in both BAM files.

I would greatly appreciate your insights on:

Which workflow would you recommend as more reliable for our analysis?
Could you help explain the potential reasons for such a significant difference in variant calling despite the similar coverage patterns?
Thank you very much for your time and assistance.

Best regards