393. Multivariate application of artificial neural networks for genomic prediction

Abstract

Phenotypic data for growth traits including birth, weaning and yearling weights and genomic data of 50K SNP markers were collected from 738 Brangus. Multivariate artificial neural networks (ANN) with 1 to 10 neurons models including the inputs from genomic relationship matrix were created and applied with the learning algorithms (Levenberg-Marquardt, Bayesian Regularization and Scaled Conjugate Gradient) and activation functions (tangent sigmoid and linear) for the analysis of growth data. Pearson correlation coefficients were used to evaluate multivariate ANN model performances about genomic prediction in datasets. In multivariate ANN models, the prediction performance of the combination of learning algorithms and activation functions changed for growth traits and there were no superior multivariate ANN models with learning algorithms and activation functions. The application of different activation functions did not make any significant difference on the genomic prediction performance of ANN models with different learning algorithms.