Inferring Novel Exons From Rna-Seq Data

Question

[toshnam]

Hi,

I wonder if I can infer novel exons in RNA-Seq.

I'm used to use TopHat and Cufflinks for read alignment and reconstruction respectively. In cuffcompare output (.gtf), I could see the class code 'I' that means a transfrag falling entirely within a reference intron. Some of the transfrags were mapped on the intronic region of refgene. Can I guess the transfrag is novel exon based on the cuffcompare output.

Is there another way for inferring novel exon in RNA-Seq?

Thanks in advance.

Answer 1

[DG]

The TopHat?Cufflinks pipeline, assuming you are not restricting it to only the provided gene models (GFF/GTF file) already infers potentially novel exons. In the isoforms file at least potentially novel isoforms have a code of J.

But, as another answer points out, be wary of these as "novel" things generally have a higher probability of being some sort of false-positive. Genomic contamination, a high proportion of improperly spliced mRNAs that haven;t yet gone through nonsense mediated decay (depending on conditions of your experiment), etc. So if you see anything really interesting you want to follow up on it cautiously.

Answer 2

[pierre.peterlongo]

Hi,

We developed a few month ago a tool (still evolving) called kissplice. This tools detects alternative splicing events from raw (illumina) ngs data. Given one or several read datasets it generates the de bruijn graph from all kmers of all datasets. Then motifs corresponding to alternative splicing events are detected and initial reads are finaly mapped one them to retreive the quantitative read coverage information per input dataset.

kissplice web site

Answer 3

[swbarnes2]

I think if your read falls entirely in an intron, you have to wonder if that's just genomic contamination, and not a novel exon. I think you would need a read that crosses from one exon to the other, to be sure that the new thing really is a spliced exon.