With the tremendous growth in biological information, bioinformatics has become a powerful approach to aid in assigning the functional role of proteins. By establishing an ancestral relationship or homology to a well-understood protein, the function of a previously uncharacterized protein can be inferred.
The most common method to detect homology between proteins is to use sequence alignment, of which BLAST and PSI-BLAST are the most popular tools. The challenge is to find as many true positives as possible, and distinguish these true positives from false positives when sequence similarity falls into the twilight zone (<25%), as is commonly observed for distantly related sequences.
A two-pronged approach is presented to address the challenge in distant homology detection. In the proposed LESTAT algorithm, conserved structural features are incorporated into an iterative profile-based sequence alignment method. This imparts LESTAT with the ability to finding more true positives than PSI-BLAST based on seven test case studies. In the proposed SimpleIsBeautiful (SIB) algorithm, a mathematical model and a novel model validation approach is utilized to improve PSI-BLAST's ability to discriminate true and false positives without sacrificing its computational efficiency. These additional features result in improved performance in deciphering true and false positives when compared to existing PSI-BLAST approach. A web-server that runs the SIB algorithm, SIB-BLAST, was launched in December 2008 under the URL (http://sib-blast.osc.edu).
One alternative application of homology prediction is to utilize that information to predict protein-protein interactions. As a first step to explore such questions, an algorithm was developed that attempts to predict interacting partners of a hetero-oligomer from a homo-oligomer using a structure-based sequence alignment strategy in conjunction with correlation analysis of amino acids pair. The prediction algorithm was applied to the human Rh proteins and the SSoPCNA proteins. The results reveal that interacting residues in a homo-oligomer do undergo mutation, presumably under evolutionary pressure, when trying to complex with another protein molecule to form a hetero-oligomer.