Weak Homology Detection by Profile-Profile Comparison

Experimentally determined protein tertiary structures are currently increasing in an enormous rate. The growth of the structure database makes structure prediction methods based on known tertiary structures more efficient. Detecting weak homology is crucial to extend applicable area of these methods. For this purpose, a profile or PSSM (position-specific score matrix) derived from multiple alignments of protein families has been proved to be quite effective. One of the most successful examples of profile-based methods is PSI-BLAST. But PSI-BLAST simply compares profiles with amino acid sequences. A more sensitive search must be achieved by comparing one profile with another profile, and several studies reported that sensitivity was improved when profile pairs were aligned [4, 5]. We combined this profile-profile comparison method and predicted secondary structure information, and succeeded in significantly improving distant homology recognition performance in comparison with PSI-BLAST.