Any Ideas To Increase Hmm Classification Accuracy Of Two Groups Of Proteins?

Question

[arronslacey]

Hi I have built a profile hmm for antifreeze proteins( http://pfam.sanger.ac.uk/family/PF08666) and am testing out antifreeze proteins and antifreeze-like proteins (http://pfam.sanger.ac.uk/family/PF13144) however they both score very similarly against the hmm model.I was wondering if anyone had suggestions like biasing the transition/emission matrices to help the model discriminate between the two a bit better. anyone had any experience with biasing hmm's?

Thanks

[Neilfws]

I suspect, given that antifreeze proteins are rather "floppy", unstructured things, that you won't see much more discrimination between these groups.