Evolutionary symbolic discovery for bioinformatics, systems and synthetic biology

Abstract

Symbolic regression and modeling are tightly linked in many Bioinformatics, Systems and Synthetic Biology problems. In this paper we briefly overview two problems, and the approaches we have use to tackle them, that can be deemed to represent this entwining of regression and modeling, namely, the evolutionary discovery of (1) effective energy functions for protein structure prediction and (2) models that capture biological behavior at the gene, signaling and metabolic networks level. These problems are not, strictly speaking, "regression problems" but they do share several characteristics with the latter, namely, a symbolic representation of a solution is sought , this symbolic representation must be human understandable and the results obtained by the symbolic and human interpretable solution must fit the available data without over-learning.