Error While Running Cegma (Geneid-Train Step)

Question

[arnstrm]

I am trying to run CEGMA on the newly assembled genome (scaffolds) and I have trouble getting past the geneid step. I ran CEGMA with default parameters, as cegma --ext -g genome.scf.fasta. The pipeline ran for about 10 hours (32 procs, 256 GB RAM) and exited giving this: CEGMA, geneid error: geneid-train did not work properly. When I investigated, I found it was geneid-train step. So, I tried to run it manually as:

$ geneid-train -v local.geneid.selected.gff local.geneid.selected.dna geneid_params
DATA COLLECTED: 298 Coding sequences containing 1311 introns
Intron model
Coding model 
Use of uninitialized value in numeric eq (==) at /data004/software/GIF/packages/cegma/2.4.010312/lib/geneid.pm line 264.
some values in Markov model with zero counts, use pseudocounts at /data004/software/GIF/packages/cegma/2.4.010312/lib/geneid.pm line 270.

Does anybody have any experience with geneid? How can I get past this step? My genome's estimated size is 745 MB and has about 574K scaffolds (643 MB total length), N50=607. Any input will greatly be appreciated!

Thanks

[keith]

Your N50 is 607 bp?!? Or did you mean 607 Kbp. If it really is 607 bp then you should probably not be doing anything else with that genome assembly.

[arnstrm]

LOL, no! It was the preliminary assembly. Later it was improved and the numbers are much better now.

Still a work in progress:

Answer 1

[arnstrm]

This has been resolved on this Trouble running CEGMA on the sample dataset!

Answer 2

[keith]

This hasn't actually been resolved yet...I'm still working on what might be going on.

Answer 3

[Francisco Camara]

Hi Keith and arnstrm,

I am not sure whether this issue has been sorted but I can try to help if it hasn't. The CEGMA pipeline was developed by from a group other than ours. However it uses geneid, which an ab initio gene prediction tool developed in our group.

I have used CEGMA quite extensively to determine the quality of the protein-coding "gene-space" of different genomes but I don think I have run into this sort of problem...

It sounds like some of your input fastas may have no sequence content but I am not sure..Could you give me access to the intermediate files including the geneid parameter file (self.param) so that I could try to figure out what's wrong?

I assume I have the latest version of geneid installed on the system (1.4.4 -check here: http://genome.crg.es/software/geneid/index.html) and/or your CEGMA installation is pointing to it.

Thanks,
Francisco Camara

[arnstrm]

Thanks, Francisco! The issue has been resolved in version >2.5 (see this post for details). Keith (as in Keith Bradnam) is one of the CEGMA developer!

EDIT: didn't realize, I was talking to geneid developer as well. I am so happy now!