Tophat2 normal alignment rate

[JJP]

Hi there, I am a complete RNA seq newbie, so let me excuse myself preventively for any stupid questions.

Recently I started working with mouse RNA seq data, after running the data though Tophat2 I got a 91.0% overall read alignment rate and a 83.1% concordant pair alignment rate.

Could somebody with more experience in the field tell me if this are good alignment values, and if not which are the normal values I should expect ?

Thanks in advanced.

[Jon_Keats]

what read length, not bad if 100mers

[JJP]

Sorry I knew I had forgotten to mention something, the reads have a length of 101.

[jp.]

Dear Jon
I have posted the same problem here;

http://seqanswers.com/forums/showthread.php?p=112098#post112098

any solution ?

what read length, not bad if 100mers

[rnastar]

I just finished running tophat on about 20 cancer samples and my concordant pair alignment rate ranged between 50-60% for 50 bp reads. Should I be worried? Will this have an impact on the downstream cufflinks/cuffdiff analysis?

[westerman]

I presume you are talking about *human* samples. The 50-60% rate is low. I suspect that since the samples are more-or-less uniform in their mapping rate then you won't have too many problems with cufflinks/cuffdiff aside from not having as much data to work with.

Did you run non-cancer controls? Or are those part of the "20" you mention above. How about replicates -- technical and biological?

[rnastar]

Yes, human samples. We did run non-cancer controls but they did not have this problem. What do you think this means? Is this more a reflection of biology or an indication of a technical issue? We do not have any technical replicates, only biological.

[westerman]

The lack of technical replicates is not a big concern -- although they can help troubleshooting usually the technical process is flawless. Although when problems do occur you really wish that you had technical reps. :-)

As for the question "is a 50-60% concordant rate normal for cancer samples when the non-cancer samples have a higher concordant rate?" ... I am going to have to skip on answering this. I simply do not have enough experience with human cancer samples to say one way or another. Cancer genome rearrangements can be widespread so may 50-60% if expected. Can anyone with more human experience chime in?

BTW: You may wish to make a new thread on this topic with the keywords "human"
and "cancer" in it. That might attract more knowledgeable people.