Soft-clipped alignments - keep or filter out?

[mortonm33]

I have question regarding the post-processing of RNA-Seq data. I'm using GSNAP version 07-03 for alignment of paired-end data and am getting a lot of soft-clipped alignments (>20% of reads). I'm wondering if it would be better to keep or remove these alignments before further processing, such as differential expression analysis. I ran my analysis before removing the soft-clipped alignments and validated one of the differentially expressed transcripts with qPCR, but then was told by a colleague that I should remove soft-clipped alignments as they don't represent real alignments. After filtering out the soft-clipped alignments, the transcript I validated with PCR (and several others originally present in the differential expression results) no longer shows up as differentially expressed in my analysis of the RNA-Seq data. Any advice is appreciated.

Thanks,
Matt

[DZhang]

Matt,

Run your data through FastQC to examine the quality score distribution along the read length. If you see a sharp drop at the last few bps, you may clip those few bps before mapping.

Best regards,
Douglas

https://www.contigexpress.com