HMM method for exploring divergent family relationships.
-
A hidden Markov Matrix (HMM)
is like a Psi-Blast PSSM except it also learns position specific gap penalties
during training. That means it can not only detect divergent sequences
but also align them (Eddy,
1998).
-
The Sequence Analysis and Modeling System (SAM) was ported
to Dr. Demeler's supercomputer.
-
SAM features context-specific prior knowledge, Dirichlet
mixtures, free insertion modules, and inclusion of structural information.
-
SAM was able to build a >400 member family containing all
known phage (and Herpes Virus) terminase genes.
-
The SAM alignment revealed structural similarities at the
level of predicted 2ndary structure maps.
-
SAM was automated to build families for every gene in a given
virus, hence making a library of family models for detecting deeply diverged
family members in new phage genomes.