problem with de novo assembly of ESTs

[bioben]

Hi,

I am trying to assemble ~10 million 454 ESTs and ~1 million sanger ESTs. I tried newbler, CAP3 and TGICL. They all output identical sequences more or less in the contigs and singlets files.

For example, I found the 454Isotigs.fna file contains many sequences that are 100% identical but with different lengths (i.e. one sequence contains another shorter one). Isn't this supposed not to happen. I mean they should be assembled as one?

I also tried CAP3 and TGICL. Again, they also output identical sequences more or less in the contigs and singlets files.

Does anyone know why? Thanks ...

[Jose Blanca]

Maybe you could try with Mira, but I warn you that with no assembler you'll get perfect results. Most assemblers a focused on genomic sequences (no splicing and even coverage). They can be tweaked a little to assemble transcriptomes, but most of the times the results are not nice.
Theoretically newbler has a transcriptome mode, but last time I tried, the result was quite poor. After much testing I stuck with Mira despite its notable problems with the transcriptome.

[flxlex]

The phenomenon of the almost identical sequences in the 454Isotigs.fna file is explained here and here (my blog on newbler, where I suggested running cd-hit on the isotigs for each isogroup to collapse them). At least newbler is trying...