Genomic prediction through machine learning and neural networks for traits with epistasis

Weverton Gomes da Costa; Maurício de Oliveira Celeri; Ivan de Paiva Barbosa; Gabi Nunes Silva; Camila Ferreira Azevedo; Aluizio Borem; Moysés Nascimento; Cosme Damião Cruz

doi:10.1016/j.csbj.2022.09.029

Genomic prediction through machine learning and neural networks for traits with epistasis

Comput Struct Biotechnol J. 2022 Sep 24:20:5490-5499. doi: 10.1016/j.csbj.2022.09.029. eCollection 2022.

Authors

Weverton Gomes da Costa¹, Maurício de Oliveira Celeri², Ivan de Paiva Barbosa³, Gabi Nunes Silva⁴, Camila Ferreira Azevedo³, Aluizio Borem³, Moysés Nascimento², Cosme Damião Cruz¹

Affiliations

¹ Department of General Biology, Bioinformatics Laboratory, Federal University of Viçosa, Viçosa, MG, Brazil.
² Department of Statistics, Laboratory of Computational Intelligence and Statistical Learning, Federal University of Viçosa - UFV, Viçosa, MG, Brazil.
³ Department of Agronomy, Federal University of Viçosa, Viçosa, MG, Brazil.
⁴ Department of Mathematics and Statistics, Federal University of Rondônia, Ji-Paraná Campus, RO, Brazil.

Abstract

Genomic wide selection (GWS) is one contributions of molecular genetics to breeding. Machine learning (ML) and artificial neural networks (ANN) methods are non-parameterized and can develop more accurate and parsimonious models for GWS analysis. Multivariate Adaptive Regression Splines (MARS) is considered one of the most flexible ML methods, automatically modeling nonlinearities and interactions of the predictor variables. This study aimed to evaluate and compare methods based on ANN, ML, including MARS, and G-BLUP through GWS. An F2 population formed by 1000 individuals and genotyped for 4010 SNP markers and twelve traits from a model considering epistatic effect, with QTL numbers ranging from eight to 480 and heritability ( $h^{2}$ ) of 0.3, 0.5 or 0.8 were simulated. Variation in heritability and number of QTL impacts the performance of methods. About quantitative traits (40, 80, 120, 240, and 480 QTLs) was observed highest R2 to Radial Base Network (RBF) and G-BLUP, followed by Random Forest (RF), Bagging (BA), and Boosting (BO). RF and BA also showed better results for traits to $h^{2}$ of 0.3 with $R^{2}$ values 16.51% and 16.30%, respectively, while MARS methods showed better results for oligogenic traits with $R^{2}$ values ranging from 39,12 % to 43,20 % in $h^{2}$ of 0.5 and from 59.92% to 78,56% in $h^{2}$ of 0.8. Non-additive MARS methods also showed high R2 for traits with high heritability and 240 QTLs or more. ANN and ML methods are powerful tools to predict genetic values in traits with epistatic effect, for different degrees of heritability and QTL numbers.

Keywords: Genome wide selection; Genome-enabled prediction; Multivariate adaptive regression splines; Non-additive effects; Quantitative trait locus.