Tophat and deep coverage question

[mathew]

I have been doing RNA-seq experiments and noticed an unusual out put of Tophat; I used two datastes of PE X50 reads for human. Top one is deep sequence of around 250 million reads while the bottom is regular 25 million reads. If you notice in the top fig there are number of reads which align to human genome but are not corresponding to exons/transcripts. While the low coverage (bottom) provide an excellent robust alignments to transcripts perfectly corresponding to cuff Ids ( refer to Fig).
The questions is how to explain the area where there are lot of reads not corresponding to transcripts/exons (marked with black arrow). I sis that TopHat cannot handle such deep coverage or it is noise? Any feedback is appreciate.
Thanks

[pbluescript]

It's hard to know what is going on without more information, but it probably has nothing to do with Tophat. For example, it could be genomic DNA contamination or it could be real and indicate retained introns, erroneous transcripts, unspliced RNA, a microRNA, etc. Many of these things might not be that well represented in a sample, so you would need deep coverage to see them.

[mathew]

details

I can share all the details what you need. Please advise what information may be helpful. I can send an email to you if you agree.

Thanks.