Many sequences, with low percentages of sequence identities - in the twilight zone - or even with statistically insignificant sequence similarities, have similar topologies and often similar functions. Recognition of topology from sequence may therefore give rise to testable hypotheses about function. Such analyses are probably best achieved at the level of globular protein domains, which are compact folding units often associated with specific functions in multi-modular proteins.
We have established a database (DDBase) of globular domains using an automated algorithm, DIAL, which identifies elements of secondary structures that form compact clusters. The protein domains have themselves been clustered on the basis of 3-D structures, using methods such as COMPARER and SEA, into superfamilies with little sequence similarity but similar topologies and functions. The lecture will focus on examples where our group has contributed through its experimental programme, using X-ray crystallography, for example the cystine knot super-family of growth factors and hormones, adaptor domains of signal transduction, the elaborated jelly-roll lectin superfamily, aspartic proteinases and matrix metalloproteinases.