Genomic prediction of leaf rust resistance to Arabica coffee using machine learning algorithms

Genomic selection (GS) emphasizes the simultaneous prediction of genetic effects thousands scattered markers over genome. Several statistical methodologies have been used in GS for merit. In general, such require certain assumptions about data, as normality distribution phenotypic values. To circumvent non-normality values, literature suggests use Bayesian Generalized Linear Regression (GBLASSO). Another alternative is models based on machine learning, represented by Artificial Neural Networks (ANN), Decision Trees (DT) and related possible refinements Bagging, Random Forest Boosting. This study aimed to DT its predicting resistance orange rust Arabica coffee. Additionally, were identify importance characteristic interest. The results compared with those from GBLASSO ANN. Data coffee 245 plants genotyped 137 used. presented equal or inferior values Apparent Error Rate obtained DT, GBLASSO, Moreover, able important Out 14 most analyzed each methodology, 9.3 average regions quantitative trait loci (QTLs) disease listed literature.