Spliced aligner for 454 reads?

[sulicon]

Hi all,
I am trying to assemble 454 reads by aligning them to reference genome.

I have tried the official gsMapper before, which just gave me exon sequences, rather than transcripts sequences. Now, I am using BWA-SW to align the reads, and Cufflinks/Scripture to reconstruct the transcripts. Still, I can only get the individual exon sequences, instead of expected transcript sequences.

It seems the problem here is that BWA-SW is not a spliced aligner so that the splicing junction reads will be lost during the alignment. A spliced aligner, Tophat, is used in the tutorial of both Cufflinks and Scripture. However, this aligner is based on Bowtie, an aligner designed for short reads only.

Could anyone give me some suggestion of the spliced aligner suitable for 454 reads? I think BLAT would be an option, but still want to test some other methods developed recently.

Thanks,
Shuli

[lh3]

For a read bridging a splice junction, bwasw should give two or more hits unless one of them is too short. Perhaps Cufflinks is expecting some tophat/bowtie specific information to group local hits to a transcript. I do not know.

Nonetheless, I agree bwa-sw would not work well because it is a local aligner. For RNA-seq/ESTs, a dedicated splicing-aware glocal aligner is more appropriate. In addition to blat, you may also try gmap.

[Jose Blanca]

gmap can align 454 ESTs against a genome taking into account the introns.

[jochensupper]

Hi,

our mapper (Genomatix) has a local spliced alignment mode that allows to align complete transcripts to the genome. Attached is a screenshot of assembled (velvet) RNA-Seq reads mapped to the reference genome.

Depending on the organism and your objective you could also consider mapping your reads against a transcriptome library (with no worries about splicing your reads). Then, however, you wouldn't be able to discover novel transcript variants.

[sphil]

Blat is also able to handle 454-splice reads but be aware of long runtime....

[sulicon]

Thanks Li Heng and Jose. I have tried GMAP and checked the output. It looks good. According the a document I have found, GMAP outperforms BLAT for gene structure identification in both speed and accuracy. And the latest version of GMAP can generate output in SAM format directly, which facilitates the subsequent analysis.

Finally, I got some transcripts constructed by running Cufflinks on GMAP output. However, it seems that Cufflinks has modified the original alignment and generated some artificial exons/splicing junctions I've never seen in the GMAP output...

[brdido]

However, it seems that Cufflinks has modified the original alignment and generated some artificial exons/splicing junctions I've never seen in the GMAP output...

Maybe this is why (from http://cufflinks.cbcb.umd.edu/manual.html):

-g/--GTF-guide <reference_annotation.(gtf/gff)> : Tells Cufflinks to use the supplied reference annotation (GFF) to guide RABT assembly. Reference transcripts will be tiled with faux-reads to provide additional information in assembly. Output will include all reference transcripts as well as any novel genes and isoforms that are assembled.

I'll try to use gmap+cufflinks for 454 data! Thanks.