STAR should sjdbOverhang be the same for every sample in a differential analysis?

[Jonathan Sage]

I know that sjdbOverhang should be set to min(100, max_read_length -1). However, if I am running STAR on multiple samples for use in a differential analysis, should sjdbOverhang be set the same for every sample in the analysis (regardless of whether the samples are composed of read files with differing max_read_length)?

[Alexander Dobin]

Hi Jonathan,

ideally, for all types of analyses, the sjdbOverhang should be adjusted for every sample according to the read length.

This would minimize bias for mapping to annotated junctions.

However, in reality the effect will be negligible, especially compared with other DE biases introduced by the varying read length.

Cheers

Alex

[Jonathan Sage]

Perfect, that's how I have it set up. Thanks for the quick reply.