The yeast genome contains 6270 genes, whose protein products range from about 25 amino acids to nearly 5000. Proteins fold into 3-dimensional structures, and the function is dependent on this structure. Larger proteins, for which a structure is known, display a modular design where parts of the amino acid chain fold into an independent structure, a domain. Statistics from the protein data bank show that the average length of a domain is about 163 amino acids for all-alpha or all-beta proteins. Clearly, numerous proteins in yeast have more than one domain.

Often, a specific function can be ascribed to a single domain, and domains of the same amino acid chain can have different functions. Hence, to describe a function of a protein, its domain configuration has to be known. To identify these domains from a sequence is far from straightforward. Our lab has developed a domain-parsing algorithm, GINZU, that makes use of several computational methods and databases to divide the target protein into domains. GINZU also assigns structural homologues to domains when the domain is identified by PSI-BLAST (Position-Specific Iterated blast) or ORFeus, a threading/fold recognition server.

The domains that have no structural homologue are subjected to ab initio protein folding with Rosetta, which most often results in a putative SCOP (Structural Classification of Proteins) superfamily assignment. Rosetta is one of the best ab initio algorithms today, as has been demonstrated in the Critical Assessment of Techniques for Protein Structure Prediction (CASP).

CASP || GINZU || Rosetta || Examples