ClipBam causing strand bias

Hi all,

I have been using fgbio a while and have been using it for a number of our lab’s cfdna analyses. My concern is that the ClipBam function might be causing some strand bias when it comes to variant calls.

I have a purchased cell line with a spiked in variant at 5%. After aligning, grouping, calling concensus reads and aligning those concensus reads I clip the over lapping reads prior to variant calling with:

java -jar fgbio-0.6.1.jar ClipBam \
    -i bam \
    -o sort.bam \
    -m clipbam.metrics \
    -r ref_fasta \
    --clip-overlapping-reads=true \
    -c Hard 

And I end up with a 2:1 strand-bias in the bam

Compared with the same sample but without ClipBam

If there is a variant on both strands, how does ClipBam decide which starnd to keep the variant on?

Any help would be really appreciated.

Thanks

Lewis

[PedalheadPHX]

exome capture does create a strand bias on the edge of the two edges capture target, so if your variant is to one side of the capture targets you can see a strand coverage bias. For specific targets you can avoid this is capture design by increasing the capture window so the edge bias of the capture is outside your area of interest

[Lewis-Gallagher]

Thanks @nh13 ! I guess it explains it a bit, though I do see that in the middle of exons as well, which in theory shouldn’t be affected by this. But then these might be due to pure chance, when we are talking about a few reads