[ssml] alignment of membrane proteins

Kevin Karplus karplus at soe.ucsc.edu
Thu Aug 19 11:23:31 EDT 2004


HMMs work just as well on membrane proteins as they do on globular proteins.
Profile HMMs are are quite effective at identifying and aligning
GPCRs, for example, and one of the best transmembrane helix annotation
tools is TMHMM.

The quality of an HMM is very dependent on the number and diversity of
the sequences it is trained on and on details of the training procedure.
If you take a poor multiple alignment and build a profile HMM from it,
you will get a poor HMM.  If you train an HMM on sequences that are
not part of the family, you will get a poor HMM.  If you train an HMM
on only a small subfamily, you will get an HMM that does not
generalize well to the full family.  All these caveats apply equally
to globular and membrane proteins, but there is less structural data
available for membrane proteins, so people are often tempted to push
the tools into much more distant relationships, often with poor results.

Kevin Karplus 	karplus at soe.ucsc.edu	http://www.soe.ucsc.edu/~karplus
Senior member, IEEE	Board of Directors, ISCB (starting Jan 2005)
Professor of Biomolecular Engineering, University of California, Santa Cruz
Undergraduate and Graduate Director, Bioinformatics
Affiliations for identification only.





More information about the ssml-general mailing list