REMOTE HOMOLOGY DETECTION WITH HMMs AND STRUCTURAL ISSUES

Computational methods for homology detection between protein sequences have become a central component in genome analysis. Nowadays, sequences of unknown function are routinely searched against databases of known proteins, providing an important aid for sequence annotation and for guiding laboratory experiments. Although homology identification through pairwise sequence matching [1, 2] is still an important tool, this approach can only detect homologue proteins that exhibit significant sequence similarity. Profile hidden Markov models (HMMs) [3,4] consider information from a number of sequences, and are known to perform better than pairwise methods in detecting weak or remote homologies [5, 6].