Solving large protein secondary structure classification problems by a nonlinear complementarity algorithm with {0, 1} variables

摘要

The aim of this paper is to present a nonlinear complementarity algorithm, limited to binary variables, to implement a classification algorithm, which will be applied to the determination of the secondary structure of proteins, and to present its statistical and mathematical convergence properties. Extensive application results will be given on the available test data sets of proteins and the results will be compared with those of other implementations. The increase in the recognition accuracy, which will be in evidence, will be shown to be attributable to the adoption of a strict statistical methodology, which will permit to obtain unbiased results, without requiring extraneous assumptions.