Protein Homology Search using Hidden Markov Model Parameters and Genetic Algorithms

S.F. Smith (USA)


Bioinformatics, homology, database search, genetic algorithms


Hidden Markov models of protein domain families are very powerful descriptions for use in protein database searches. The ability of these models to incorporate position-specific insertion and deletion probabilities as well as position-specific amino-acid substitution information gives more detail than non-position-specific methods such as Smith-Waterman or BLAST which are based on standard amino-acid substitution matrices. The drawback for protein database search is that database scoring using dynamic programming and HMM parameters is quite slow, especially when compared to using a protein domain consensus sequence in BLAST. This work proposes a method to search for protein domain family homologs using HMM parameters where the search employs genetic algorithms rather than dynamic programming.

Important Links:

Go Back