The present invention concerns methods and systems for predicting the
biological function(s) of proteins. The invention is based on the
development of functional site descriptors for discrete protein biological
functions. Functional site descriptors are geometric representations of
protein functional sites in three-dimensional space, and can also include
additional parameters, for example, conformational information. Following
their development, one or more functional site descriptors (for one or
more different biological functions) are used to probe protein structures
to determine if such structures contain the functional sites described by
the corresponding functional site descriptors. If so, the protein(s)
containing the functional site(s) are predicted to have the corresponding
biological function(s). In preferred embodiments, a library of functional
site descriptors is used to probe inexact protein structures derived by
computational methods from amino acid sequence information to predict the
biological function(s) of such sequences and of the gene(s) encoding the
same.