I'm using Pfam HMMs to identify domains in my T. cruzi proteins with the HMMER program.
I kinda understand that a HMM is build by a family multiple sequence alignment, by calculating probabilities of states.
But how the hmmscan works to classify a new protein by families HMMs? Where can I read more about this? Academic references would be appreciated.
In this context, HMMs are numerical representations of sequence alignments that are more information-rich because both residue substitution patterns and gap penalties are treated in probabilistic fashion. Diverse sequences are given higher weight in HMMs, and they also include background residue probabilities in a way that is inversely proportional to the alignment depth.
If you don't have access to Krogh, Durbin & Eddy book, the PDF version is available here. Any of the early Krogh, Eddy and Karplus papers would be good as starting points.
https://www.ncbi.nlm.nih.gov/pubmed/8744772
https://www.ncbi.nlm.nih.gov/pubmed/9918945
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
Have you tried Googling? The Durbin book should have a chapter on HMMs and HMM based classification is a relatively common topic. You could start with HMMER paper's reference section and work backwards until you get to an explanation of how and why it's done.