Mutation bias in prokaryotes varies from extreme adenine and thymine (AT) in obligatory endosymbiotic or parasitic bacteria to extreme guanine and cytosine (GC), for instance in actinobacteria. GC mutation bias deeply influences the folding stability of proteins, making proteins on the average less hydrophobic and therefore less stable with respect to unfolding but also less susceptible to misfolding and aggregation. We study a model where proteins evolve subject to selection for folding stability under given mutation bias, population size, and neutrality. We find a non-neutral regime where, for any given population size, there is an optimal mutation bias that maximizes fitness. Interestingly, this optimal GC usage is small for small populations, large for intermediate populations and around 50% for large populations. This result is robust with respect to the definition of the fitness function and to the protein structures studied. Our model suggests that small populations evolving with small GC usage eventually accumulate a significant selective advantage over populations evolving without this bias. This provides a possible explanation to the observation that most species adopting obligatory intracellular lifestyles with a consequent reduction of effective population size shifted their mutation spectrum towards AT. The model also predicts that large GC usage is optimal for intermediate population size. To test these predictions we estimated the effective population sizes of bacterial species using the optimal codon usage coefficients computed by dos Reis et al. and the synonymous to non-synonymous substitution ratio computed by Daubin and Moran. We found that the population sizes estimated in these ways are significantly smaller for species with small and large GC usage compared to species with no bias, which supports our prediction.

BACKGROUND: In structural genomics, an important goal is the detection and classification of protein-protein interactions, given the structures of the interacting partners. We have developed empirical energy functions to identify native structures of protein-protein complexes among sets of decoy structures. To understand the role of amino acid diversity, we parameterized a series of functions, using a hierarchy of amino acid alphabets of increasing complexity, with 2, 3, 4, 6, and 20 amino acid groups. Compared to previous work, we used the simplest possible functional form, with residue-residue interactions and a stepwise distance-dependence. We used increased computational resources, however, constructing 290,000 decoys for 219 protein-protein complexes, with a realistic docking protocol where the protein partners are flexible and interact through a molecular mechanics energy function. The energy parameters were optimized to correctly assign as many native complexes as possible. To resolve the multiple minimum problem in parameter space, over 64000 starting parameter guesses were tried for each energy function. The optimized functions were tested by cross validation on subsets of our native and decoy structures, by blind tests on series of native and decoy structures available on the Web, and on models for 13 complexes submitted to the CAPRI structure prediction experiment. RESULTS: Performance is similar to several other statistical potentials of the same complexity. For example, the CAPRI target structure is correctly ranked ahead of 90% of its decoys in 6 cases out of 13. The hierarchy of amino acid alphabets leads to a coherent hierarchy of energy functions, with qualitatively similar parameters for similar amino acid types at all levels. Most remarkably, the performance with six amino acid classes is equivalent to that of the most detailed, 20-class energy function. CONCLUSION: This suggests that six carefully chosen amino acid classes are sufficient to encode specificity in protein-protein interactions, and provide a starting point to develop more complicated energy functions.

The performance of two second-generation supercomputers, the NEC SX-3 and the Fujitsu VP2600, is analyzed using the Standard Los Alamos Benchmark Set, the Mendez Fluid Dynamics Codes, and some highly vectorizable production-type codes from Los Alamos.For comparison, data are also given for a single processor of the CRAY Y-MP8/264.Factors affecting performance such as memory bandwidth, vector register organization, and the effects of multiple vector pipelines are examined.On a highly vectorizable code that can take advantage of multiple vector pipes, the SX-3 and VP2600 are faster than the single CRAY Y-MP processor by factors of seven to eight.

The data transfer capacity of a vector machine is a critical factor influencing the performance of the machine. The capacity is a function of several hardware and software parameters: the bandwidth of the crossing network between processing unit and memory, the number of independent memory modules, memory access configuration, types of vector operations, and the number of access ports in-use. Memory contention occurs in a multiple access port vector machine. When long vector operation performed, the machine efficiency is just the same as the efficiency of data transfer between memory and processing unit. A random bank choice model is proposed to analyze the memory contention and its influence on machine performance. The data obtained by the model are compared with the test results on the Cray X-MP and Y-MP, and the Ardent TITAN graphic minisupercomputers. The comparisons show that the modeling data and the test results match very well. One interesting conclusion from this model is that to preserve a constant level of memory access efficiency the number of independent memory banks should be proportional to the clocks of a memory active cycle as well as to the number of access ports.

The authors present a performance evaluation and comparison of the Hitachi S-820/80 supercomputer on a set of standard Fortran benchmark codes that range from simple kernels to fluid dynamics applications. They find that the S-820 is a great deal faster in vector mode than any other supercomputer they have measured, with almost twice the performance on highly vectorized codes than the fastest machine they have seen up to now, the NEC SX-2. In scalar mode, however, the S-820 is roughly even with the X-MP and the SX-2, with a slight advantage going to the SX-2 in the applications tested.>

A partial summary of preliminary benchmark results on Japanese and American supercomputers is given. While the results indicate similar scalar capabilities between the Fujitsu VP-200 supercomputer and the one-CPU XMP, they give the Fujitsu VP-200 a clear edge over the one-CPU XMP in vector performance.