Massively parallel computation has been a significant portion of the past year's research effort. The MasPar MP-1 has been upgraded to a 16,384 processor MP-2 with a 6 Gigaflop peak performance capability. Several algorithms have been developed and adapted to the MasPar including the newly developed genetic algorithm, a very fast version of the suboptimal dynamic programming algorithm for RNA structure prediction, a very fast and sensitive sequence comparison algorithm for determining sequence homologies in proteins and nucleic acids, and a visual docker for docking drugs with a protein substrait. In addition, work has continued on the heterogeneous RNA structure analysis system with improvements in its graphical presentation capabilities, RNA database matching facilities, mutated structure generation and extensions to the MasPar interface. The above system in conjunction with gel shift experiments has been used to help determine the binding site of nucleocapsid protein NCp7 of HIV-1 and the RNA structural components that determine this site. This protein is important for encapsidation of the virus genome, RNA dimerization and primer tRNA annealing in vitro. The system is also being used to help determine the structure of human cytochrome mRNA which contains internal """"""""inframe"""""""" UGA codons of unique dual function: termination of translation or insertion of selenocysteine. Unique RNA structures have been found in the 3' untranslated region of the molecule. A new algorithm for discovering motifs in protein sequences has been developed and is undergoing testing. It uses the concept of developing the motifs from a small sample of database sequences and then refining these motifs by running them against sequences in the database. Experimental results of running this algorithm on three protein families is giving good performance.
