Computer analysis of complete prokaryotic genomes shows that microbial
proteins are in general highly conserved - similar to 70% of them con
tain ancient conserved regions. This allows us to delineate families o
f orthologs across a wide phylogenetic range and, in many cases, predi
ct protein functions with considerable precision. Sequence database se
arches using newly developed, sensitive algorithms result in the unifi
cation of such orthologous families into larger superfamilies sharing
common sequence motifs. For many of these superfamilies, prediction of
the structural fold and specific amino acid residues involved in enzy
matic catalysis is possible. Taken together, sequence and structure co
mparisons provide a powerful methodology that can successfully complem
ent traditional experimental approaches.