HMMER 3 Dirichlet mixture parameters

Question

[ivan.vujaklija]

Hi,

I'm trying to implement a Dirichlet mixture model for computing HMM emission probabilities, based on the paper by Sjolander et al 1996. My program seems to be working OK, but to be sure that my code is free of bugs I would like to compare it to the output of hmmbuild algorithm (HMMER 3), which uses Dirichlet mixture model for computing HMM emission probabilities as a default.

As far as I understand HMMER 3 uses a new Dirichlet prior, not the one from [[1]]? So could someone please tell me where can I get the new Dirichlet mixture parameters for HMMER 3. HMMER 3.1 manual says only that Dirichlet mixture prior parameters are "built in".

Thanks.

Ivan

[GenoMax]

Sounds like you are a programmer so you could look in the source code for HMMER while waiting for a response.

[SES]

That is a good suggestion, and I would recommend looking on github for browsing the source code.

[ivan.vujaklija]

Thanks guys. I found it.

The only thing which is a bit odd is that the nine distribution mixture coefficients differ slightly from the numbers provided in (Sjolander 1996), although the code comment says it is the Sjolander 96 coefficients. Numbers don't differ much, so it probably doesn't have any significant effect on the emission distributions.