A genetic algorithm for structure-activity relationships: software implementation
Abstract
The design and the implementation of a genetic algorithm are described. The applicability domain is on structure-activity relationships expressed as multiple linear regressions and predictor variables are from families of structure-based molecular descriptors. An experiment to compare different selection and survival strategies was designed and realized. The genetic algorithm was run using the designed experiment on a set of 206 polychlorinated biphenyls searching on structure-activity relationships having known the measured octanol-water partition coefficients and a family of molecular descriptors. The experiment shows that different selection and survival strategies create different partitions on the entire population of all possible genotypes.
Keywords
Cite
@article{arxiv.0906.4846,
title = {A genetic algorithm for structure-activity relationships: software implementation},
author = {Lorentz Jantschi},
journal= {arXiv preprint arXiv:0906.4846},
year = {2009}
}
Comments
21 pages; 10 figures; 14 tables