Training Augustus with tophat

[k-gun12]

I've got two lanes of Illumina 120x120 cDNAs and a ~70Mb denovo assembled genome. I'd like to train augustus using the .bam alignments of said transcripts to said genome. The webpage reads thusly:

In lack of a better program, we chose to use the program BLAT, which is not customized for NGS reads (Next Generation Sequencing), but found more introns than Tophat 1.0.11. However, it is straightforward to adjust this pipeline to Bowtie/Tophat input, because Tophat produces results in the same or similar format (wig file and intron list including multiplicities).

I'm a bit confused regarding this as I don't believe either .wig files or intron lists are part of the tophat output. Has anyone done this?

I have no problem using BLAT as well, however I've already spent 3 days of cluster horsepower on tophat.

Thanks!

[glacerda]

k-gun12, the information you are looking for is on accepted_hits.sam and junctions.bed (fort the intron positions)

Youl could go from sam to wig using a sam2wig script you can find on the web (I have downloaded this before, but could not remember where). The information you need to find the intron positions is in junctions.bed, but I don't know any script for conversion, in this case. However, it is not difficult to write a script to accomplish this conversion.

I would suggest to filter out junctions supported by only 1 read, as many of them are false positive.