The use of artifical neural networks for parameter determination in biological systems


In this paper an algorithm was developed for DNA state simulation using hydrogen bonds between complementary pairs and stacking interactions between neighboring hase pairs. In the aim to investigate the information transfer phenomenon between oligonucleotides, the hybridization reaction was modeled as a noisy channel in which error-free transmission of information occurs between perfect Watson-Crick complements, and erros occur if non-Watson-Crick base pairs are presented. The chemical potential of compounds was derived and a new group contribution method for DNA activity was developed. This method was implemented by using an artificial neural network. Parameters of different kinds of interactions and size differences between the molecules were predicted. An algorithm for DNA information processing, which involved hybridization equilibrium constants, activity coefficients of chain association and association constants, was develped. Melting profiles and local map stability prediction for sequenced DNA were illustrated. The amount of information that can be transmitted without error is bounded by the capacity of the channel. The main result in this paper is a new evolutionary algorithm which includes an artificial neural network model for DNA parameter determination and DNA state processing. The results of the communication depend on the hybridization reaction and reassociation, which also determine the number of molecules that can be interpreted as results without error. The computed results for hybridization energy were in a good agreement with the experimental data of other authors.

Publication Title

Chemical Industry and Chemical Engineering Quarterly

This document is currently not available here.