Heavy read stacking on the solexa platform

[srao]

Hello,

Not sure if someone already brought this up here before, but has anyone looked into the heavy stacking problem that is evident on the Solexa (and SOLiD) platforms?

From the alignments, it is obvious that some regions have very deep coverage with reads starting and ending at the exact same position (This could be the same read replicated numerous times). This seems to be a library specific issue, and cannot all be accounted for by repeats. Some libraries are worse than others, and the situation is compounded most likely at the PCR step.

Has anyone worked on normalizing such reads? A simple collapsing to a consensus might result in loss of valid snp information. Besides one should be able to distinguish between valid coverage and stacking because of a bad library.

Any thoughts?

Thanks.

[bioinfosm]

I believe its not just library, but also sequence related - fragile regions of the sequence break more readily giving high coverage there.., etc

[lh3]

This may be caused by PCR duplicates. You should ask wet-lab people to include sufficient large number of molecules before PCR. You can remove PCR duplicates with paired end reads.

Duplicates may also be caused by overlapping clusters. You can tell this from the coordinate of a read on the image.

Fragile region is an alternative cause, but to my experience, this is not the leading cause at least on resequencing.

[zee]

I just noticed this with some of my nucleosome data. With microRNA reads this is what is expected so lots of people do tag counts before mapping their tags, ofcourse expect to lose that quality information.
This week I was trying out the FindPeaks program and it has an option to automatically discard these when you're doing peak detection.

[swbarnes2]

Someone at the Illumina user's meeting said that they split their last round of apmlification into 4 parts, and that this helped improve library diversity.