UCSC Protein Structure Prediction and SAM 2.2

Richard Hughey rph at cse.ucsc.edu
Thu Nov 12 14:30:29 EST 1998


-----------------------------------------------------------------
	 UCSC SAM-T98 Protein Structure Prediction	      

http://www.cse.ucsc.edu/research/compbio/HMM-apps

We are pleased to announce the availability of a hidden Markov model
(HMM) protein structure prediction server.

The server has used the SAM/T-98 method to create a library of HMMs,
one per PDB structure (about 2500 HMMs total).  You can search this
database of HMMs with a protein sequence.  The iterative method of
creating these models is detailed in two upcoming papers available
from our WWW site (to appear in JMB (in collaboration with Jong Park,
Timothy Hubbard, and Cyrus Chothia) and to appear in Bioinformatics),
and is more sensitive for remote homology detection than PSI-BLAST or
ISS.  These methods, refinements of our CASP2 methods, formed the core
of our CASP3 (http://predictioncenter.llnl.gov/) structure prediction
contest entries, the results of which will be announced in December.

You will receive by e-mail a list of the PDB identifiers of each hit,
as well as a series of pairwise alignments based on the library's HMM
for those structures.  When the system is unloaded, the search will
take a few minutes.

Also available on the subpage are SAM-T98 database searching,
alignment comparison, and alignment refinement.

SAM-T98 database searching can take a particularly long time when the
server is processing many queries.  Please wait at least a day before
giving up on a search.

------------------------------------------------------------------
	SAM Sequence Alignments and Modeling Software Suite 2.2

http://www.cse.ucsc.edu/research/compbio/sam.html

An upgrade of the UCSC HMM software is also available (it does not
currently include the SAM-T98 method).  The software includes tools
for building HMMs from aligned and unaligned sequences using a variety
of Dirichlet mixture priors and transition regularizers, and scoring
and multiply aligning sequences using the trained HMM.  The object
code is free for academic use, but our copyright office would like a
signed license, the details of which are on the WWW page.

Important additions to recent versions include:

     o An option for posterior-decoded alignments
     o Local and semi-local training and alignment
     o User-defined alphabets
     o Reduced space dynamic programming
     o Optional internal sequence weighting during training
     o Corrected MSF and HSSP file reading


------------------------------------------------------------------
      Other UCSC Servers of Interest

http://www.cse.ucsc.edu/research/compbio




More information about the Bio-soft mailing list