Premium
Parallel implementation of a protein structure refinement algorithm
Author(s) -
Gunn John R.,
Friesner Richard A.
Publication year - 1996
Publication title -
journal of computational chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.907
H-Index - 188
eISSN - 1096-987X
pISSN - 0192-8651
DOI - 10.1002/(sici)1096-987x(19960730)17:10<1217::aid-jcc3>3.0.co;2-r
Subject(s) - computer science , algorithm , simulated annealing , minification , pairwise comparison , myoglobin , table (database) , node (physics) , artificial intelligence , data mining , chemistry , organic chemistry , structural engineering , engineering , programming language
Abstract An algorithm is presented for the refinement of reduced‐model structures of proteins. A simulated annealing minimization is carried out in which the trial moves consist of the replacement of three‐residue segments from a presorted library. The segments in the library are screened independently by their net end‐to‐end rotations soas to reproduce a distribution of conformations in the library similar to those in theensemble. A general form of a look‐up table contact potential is used to evaluate the free energy. This algorithm has been implemented on a parallel connection machine, on which a large number of molecules can be simultaneously simulated. The calculation of the pairwise distance matrix is distributed across the nodes of the machine to achieve an increase in performance and a reduction in the memory required on each node to storethe potential table. The results of the refinement are shown for the test case of myoglobin, and the parallel performance is compared to that of a serial version of the algorithm. © 1996 by John Wiley & Sons, Inc.