TopHat & Cufflinks failing to assemble full length transcripts

[jlhaner]

Hi,

First post on SeqAnswers. The discussions here are very useful.

We are using Tophat (v1.0.13) and Cufflinks (0.8.3) without a reference GTF and then use Cuffcompare to identify the assembled transcripts.

We are finding many transcripts reported as novel isoforms that we suspect are actually just the main transcript being divided into 2 fragments, or are leaving out a few exons at the beginning which are clearly covered by reads.

For example, if the gene has 34 exons, novel isoform j1 is identified as the first 23 exons, and j2 is made of the last 11 exons. Another example, a novel transcript is reported which begins from exon2, however there are an equal # of reads covering the first exon.

Examination of the .wig file shows coverage of the complete transcript but for some reason the full length transcript is not being assembled. We've changed the # of bp on either side of the splice junction, with no avail. We also run the butterfly search, and no change with that option either.

Does anyone have suggestions for us?

Thanks,
Jessica

[Cole Trapnell]

Are you certain there are spliced reads connecting those exons as well? You want to visualize the read alignments in IGV to ensure that you have reads spanning all the junctions.

[litc]

I'm also interested in mRNA-seq reads assembly. It seems that there is not a good soft do that work because of the alternative splice.

[lshen]

Try new version instead of this Cufflinks (0.8.3). I got greatly improved transcript assembling results with my single read data. Lot of them "full" length when compared with existing annotations.