A simulated annealing approach to the search problem of protein crystallography

With the growing availability of computer power it has become routine to perform exhaustive multidimensional searches in protein crystallography. Specifically, in cases where homologous or partially homologous structures are available, the initial interpretation of poor electron density maps is done by performing computer-intensive rotational and translational searches in real space. Often such calculations of the best fit between structure and map cannot even be attempted owing to the vast computing effort involved (years of MicroVax II time). Here, the combinatorial optimization method, simulated annealing, is shown to reduce substantially the computing effort involved and also to permit computations that are beyond the reach of current algorithms. This is illustrated with practical examples involving the structure determinations of the human histocompatibility antigen HLA-A2 and an influenza virus hemagglutinin-sialic acid complex.