Finding Reads That Suffered Splicing

Question

[Gabrielw]

I've aligned 454 RNA-Seq reads using bwa. Does somebody know if a software that identify variants that suffered splicing exists ?

Answer 1

[Sean Davis]

TopHat is probably not the "best" in terms of sensitivity or specificity. Have a look here for a more complete list. I use GSNAP quite regularly for this task.

[lh3]

Yes, I agree. TopHat does not do gapped alignment, while 454 has this particular homopolymer errors. TopHat may not work well for long reads, either. Perhaps even blat/gmap will be better.

[Sean Davis]

Good point. Blat/gmap should do well; I have not seen much 454 data, so I don't have a good sense of how well, though.

Answer 2

[Rm]

For RNA-seq data try Tophat; TopHat is a fast splice junction mapper for RNA-Seq reads.

Answer 3

[Cjt]

I recommend to have a look at PALMA, described here: http://www.fml.tuebingen.mpg.de/raetsch/suppl/palma

[lh3]

I used to read this paper. This figure is greatly exaggerating the problem with mapping ESTs. The authors are fabricating/selecting ESTs with extremely short exons and claim the alignment is correct only if all splice sites are correct. For most ESTs and when measuring the exon-level accuracy, the difference between PALMA and BLAT should be negligible (a couple percent at most). PALMA relies on BLAST, which is very inefficient for human alignment.

[lh3]

I used to read this paper. This figure is greatly exaggerating the problem in mapping ESTs. The authors are fabricating/selecting ESTs with extremely short exons and claim the alignment is correct only if all splice sites are correct. For most ESTs and when measuring the exon-level accuracy, the difference between PALMA and BLAT should be negligible. PALMA relies on BLAST, which is very inefficient for human alignment.