Parallel computing system for the efficient calculation of molecular similarity based on negative electrostatic potential
Theoretical Chemistry Group, Engineering Faculty, Universidad Nacional de Colombia, Bogota, Colombia
Universidad Nacional de Colombia, 2011
@article{torres2011parallel,
title={Parallel computing system for the efficient calculation of molecular similarity based on negative electrostatic potential},
author={Torres Carvajal, R.E. and others},
year={2011}
}
This document proposes an alternative method for the comparison of molecular electrostatic potential (MEP), based on parallel computing algorithms on graphics cards using NVIDIA CUDA platform and kernel methods for pattern recognition. The proposed solution optimizes the construction process of a particular representation of MEP, presents options for improving this representation, and offers 11 kernel functions for comparison process. Experimental evaluation using cluster analysis shows good results on a set of 73 molecules from classic chemistry functional groups, finding relationships of acidity and basicity using edit distance and detecting presence of oxygen, nitrogen and functional groups (C=O or C-O) when kernel functions are used.
December 29, 2011 by hgpu
