The Parallel Strategy Research Of Drug Compound Similarity Computing

The traditional drug discovery process is not only costly, but also needs longer period and has higher risk. The development of computer-aided drug design technologies especially virtual screening has greatly improved the efficiency of drug discovery. The most commonly used method in virtual screening techniques is to calculate the molecular similarity. Multi-core CPU and GPU technology development provides a convenient to rapid screening of mass compounds.The paper mainly studies the parallelization strategy of two molecule similarity computing methods. Among the molecule similarity computing methods, the first one is based on Gaussian shape, the other is based on molecular shape and chemical characteristics. For the first method, implement the algorithm. Propose and realize a parallelization strategy that rely on CPU and GPU collaborative computing and a strategy totally based on GPU. Compare and analyze the parallel performance of the various strategies. For the latter, realize a accelerated strategy, and propose a GPU parallelization strategy on the basis of previous experiments. Taking into consider of the molecular and atomic data independence in the molecule similarity computing. Sum up three types of parallel methods of solving problem like calculating molecular similarity on the basis of the two experiments results. Analyse the scope of application for each parallel method.