The difference between the number of proteins with known sequence and those with wellstudied function (sequence-function gap) is growing daily. Bioinformatics bridges this gap mostly through homology-based inferences of the type: assume that the function of A is known, and that B is sequence similar to A. The inference then is that A and B both have the same function. Successful homology-based inference requires that function is conserved at the level of divergence between A and B, and that an alignment between the two picks up the correct signal of conservation. Here, we propose the development of a new generalized alignment method that will be tailored specifically to the identification of competitive protein-protein binders, e.g. proteins A and B that bind well to the same protein target. Such a development is particularly needed in the light of the recent finding that protein-protein interactions are poorly conserved between organisms. Our main hypothesis is that competing binders will have similar binding hot spots and that the accuracy of our method for the prediction of hot spots will suffice to profit considerably from focusing exclusively on such hot spots when comparing two proteins. In other words, we will first predict the hot spots for both A and B and then align A and B preferentially through their hot spots. While no such concept has ever been applied to the prediction of binding features, similar concepts have significantly improved the identification of proteins with similar structure.
