ANALYSIS OF PROTEIN FAMILY CLUSTERS

Computational Biology Centers, Academic Health Center, University of Minnesota, Box 43 Mayo, 420 Delaware St SE, Minneapolis, MN 55455 USA

Current methods for clustering protein sequences into families are a valuable tool for identifying function. There are nearly a dozen such classifications in widespread use. Each was created with a different objective in mind, and can be used to characterize unknown sequences in subtly different ways. We discuss preliminary results of a set-theoretic comparison and practical-utility study of several publically-accessible databases (PIR1, DOMO, ProDom, Pfam, PROSITE, SBASE, BLOCKS, PRINTS, SYSTERS, and PROTOMAP). In comparing such a diverse set, we focus on overlap/consensus, supersets/subsets, and strengths/weaknesses for various research objectives. We do not attempt to assess 'winners' and 'losers', given the varied purposes for which each database was created.