CEGMA with custom set of proteins

[retardia]

Hi there,

I am trying to use CEGMA with a custom set of alignment that I maintain, instead of the default KOGs. When running in a custom mode, one has to provide HMM profiles created with HMMER and "choose a cutoff for each profile". Example from the CEGMA readme:

cegma --genome sample.dna --prot_num 4 --protein ORTH.fa \\
--hmm_prefix ORTH --hmm_profiles hmm_profiles/ \\
--cutoff_file profiles_cutoff.tbl

I have 2 questions:

1. Anyone knows what "the cutoff for the HMMER alignments" refers to, and how the default profiles_cutoff.tbl file was generated?

1. Do I understand it right that all alignments need to have the same number of sequences, in order to satisfy the —prot_num argument?

Any tips would be greatly appreciated, many many thanks.

Fabien

[ssully]

Hi there,

I am trying to use CEGMA with a custom set of alignment that I maintain, instead of the default KOGs. When running in a custom mode, one has to provide HMM profiles created with HMMER and "choose a cutoff for each profile". Example from the CEGMA readme:

cegma --genome sample.dna --prot_num 4 --protein ORTH.fa \\
--hmm_prefix ORTH --hmm_profiles hmm_profiles/ \\
--cutoff_file profiles_cutoff.tbl

I have 2 questions:

1. Anyone knows what "the cutoff for the HMMER alignments" refers to, and how the default profiles_cutoff.tbl file was generated?

1. Do I understand it right that all alignments need to have the same number of sequences, in order to satisfy the —prot_num argument?

Any tips would be greatly appreciated, many many thanks.

Fabien

I'd like to revive this thread because it was never answered...and I have the same questions ;>

[mendezg]

I'll revive this again since neither of you got any answers. I came across this when looking for help with a different cegma question.

The cutoff values are calculated as described in their paper. For each gene they took their alignment (of 6 taxa) and created every sub alignment of n-1 taxa. They created hmms from the n-1 alignments. Then they did an hmmsearch of the missing taxa with the hmm. They took the score of each of those searches and averaged them and then divided by 2 to generate the cutoff value.

I created a shell script to generate these scores, if anyone is interested.

The prot_num argument does mean that you will need all the alignments to have the same number of sequences. I wrote a shell script to bin all the genes based on prot_number. Another script then runs cegma with each binned prot_number.

Another issue that comes up is that the hmm used in hmm_profiles needs to be generated by hmmer 2.3.2 and calibrated using hmmcalibrate. If you use an hmm generated by hmmer 3 you get an error.