(E) Augustus (slow, needs A)

[JakubNvk]

• Run gene finder Augustus, create track with predicted genes (for comparison with annotation)
• Download and install software from http://bioinf.uni-greifswald.de/augustus/ (Already installed on compbio servers)
• Example of command line: augustus --uniqueGeneId=true --species=ustilago_maydis genome.fa > augustus.gtf ustilago_maydis is a related fungal species used for training parameters
• The result needs to be converted from gtf to genepred, by gtfToGenePred (at genomika server) with option -genePredExt
• If you name your track augustus, genome browser will recognize it automatically, no need to modify trackDb.ra

[rasto2211]

Used ustilago_maydis model (fungi) since there's no MalSym model. From augustus docs:

Note that for closely related species usually only one version is necessary. For example, the human version is good for all mammals.

[rasto2211]

Output from gtfToGenePred

augustus.gtf doesn't appear to be a GTF file (GFF not supported by this program)

[rasto2211]

It seems that we need to remove lines which have transcript and gene in the third column.

cat augustus.gtf | grep -v "^#" | grep -Pv "\t(gene|transcript)\t" > augustus2.gtf

Transcript and gene lines have data in the 9th column in a format which does not agree with GTF spec.

[rasto2211]

Convert GTF to GenePredExt:

gtfToGenePred -genePredExt augustus2.gtf augustus.genePredExt

Load data from GenePredExt file to augustus SQL table:

hgLoadGenePred -genePredExt malSym1 augustus augustus.genePredExt

[rasto2211]

No need to modify /kentsrc/trackDb/malSym/malSym1/trackDb.ra since augustus track is already defined in /kentsrc/trackDb/trackDb.ra.

[izri16]

@rasto2211 I suppose that it would be better to write those approaches to wiki rather than store them here.

[rasto2211]

I looked at augustus track in genome browser and it seems quite reasonable.

I also added docs to github wiki.