Evolving Toxicity Models using MultigeneSymbolic Regression and Multiple Objectives

Abstract

In this contribution a multi-objective genetic programming algorithm (MOGP) is used to perform symbolic regression. The genetic programming (GP) algorithm used is specifically designed to evolve mathematical models of predictor response data that are "multigene" in nature, i.e. linear combinations of low order non-linear transformations of the input variables. The MOGP algorithm simultaneously optimizes the dual (and competing) objectives of maximization of 'goodness-of-fit' to data and minimization of model complexity in order to develop parsimonious data based symbolic models. The functionality of the multigene MOGP algorithm is demonstrated by using it to generate an accurate, compact QSAR (quantitative structure activity relationship) model of existing toxicity data in order to predict the toxicity of chemical compounds.

Genetic programming for experimental big data mining: A case study on concrete creep formulation

“…The high-level crossover allows the exchange of one or more genes with another selected individual subject to the G max constraint. If an exchange of genes results in any individual containing more genes than G max , the genes will be randomly selected and deleted until the individual contains G max genes [16][17][18][19].…”

“…The algorithm attempts to maximize diversity by ensuring that no individuals contain duplicate genes. The genes are randomly chosen, and the vector of unknown coefficients d is estimated using least squares normal equation as follows [16][17][18][19]:…”

“…In contrast to standard symbolic regression, MGGP allows the evolution of accurate and relatively compact mathematical models. Even when large numbers of input variables are used, this technique can automatically select the most contributed variables in the model, formulate the structure of the model, and solve the coefficients in the regression equation [16][17][18][19]. Therefore, unlike other techniques such as traditional regression analysis or ANN, there is no need in the MGGP technique for the user to pre-define the formulation structure of the model or select any existing form of the relationship for optimization [3,4], which makes it more practical for complex optimization problems.…”

Group Contribution Method-based Multi-objective Evolutionary Molecular Design

“…Lin et al studied the design of metal catalysts [15], numerous articles can be found for the design of drugs [5,[16][17][18], Perdomo et al designed and improved biodiesel fuel blends [3] and Kasat et al summarized the applications of genetic algorithms in polymer science including polymer design [2].…”

“…The crossover Eqs. (16)(17) and mutation Eqs. (18)(19) operators are applied to create the next generation (with N members) [27].…”