Issue with Tophat and Bowtie2

[scottdson]

So I've been using Tophat (running bowtie2) for various RNA-seq mapping projects and using default Tophat settings will get more than 75% mapping on first instance, but i've came across a strange scenario i was hoping to get input on.
We got a library sequenced by strand specific RNA-seq, the sequencing centre messed up and did 100bp reads instead of 150bp reads, and agreed to re-sequence at 150bp for us. In the mean time i mapped the 100bp reads and as usual got more than 75% reads mapped with default settings. On getting the same sample re-sequenced at 150bp the mapping came back at only 38% mapped. Alarm bells started ringing that some thing was up. I thought perhaps allowing more mismatches from the Tophat default will allow more reads to map; 3 mismatches -> 41% mapped, 4 mismatches -> 44% mapped.

Has anybody came across a situation where a second sequencing of the same library has lower mappability?

[GenoMax]

Have you done QC with the longer reads set? Does that data set look as good as your 100 bp reads (no adapter contamination, short inserts etc)?

[scottdson]

Hi Genomax,

I use FastQC, and the unusual things about the dataset are that on the per base quality page the box-Whisker plot tails dip below a Phred score of 28 from 135 -151 bp, and on the Kmer content page there are some 5 mer sequences over represented from the 130 bp mark.

Perhaps I should only be mapping with the first 135 bp.

[GenoMax]

Can you try to scan your sequences to see if you have adapter contamination? Instead of using a hard cut-off of 135 bp, BBDuk may be simplest to use: http://seqanswers.com/forums/showthread.php?t=42776

It is possible that you have a population of inserts that are shorter than you expect and that is what is causing the low mapping.

[scottdson]

Thanks for the link GenoMax, this hard cut-off wasn't the most desirable option but so you're suggestion might be a very good option.
Looking back into the sample prep, the library might have been sonicated specifically for the 100 bp sequencing, so getting the re-sequencing would have been done with short inserts.

Thanks so much for the replys.