Comparative Genomic Prediction of Resistance to Fusarium Wilt (Fusariumoxysporum f. sp. niveum race 2) in Watermelon: Insights from Parametric and Machine Learning Approaches

Abstract

Complex traits influenced by multiple genes pose challenges for marker-assisted selection (MAS) in breeding. Genomic selection (GS) is a promising strategy for achieving higher genetic gains in quantitative traits by stacking favorable alleles into elite cultivars. Resistance to Fusarium oxysporum f. sp. niveum (Fon) race 2 in watermelon is complex and polygenic with moderate heritability. This study evaluated GS as an alternative or additional approach to quantitative trait loci (QTL) analysis/marker assisted selection (MAS) for enhancing Fon race 2-resistance in elite watermelon cultivars. Objectives were to: 1) assess the accuracy of genomic prediction (GP) models for predicting Fon race 2-resistance in F₂ (Pop I) and recombinant inbred line (RIL) (Pop II) populations, 2) rank and select families in each population based on genomic estimated breeding values (GEBVs) for developing testing populations, and 3) verify if major QTL associated with Fon race 2-resistance are present in top selected families with the highest GEBV. Resistance ratings were based on the percentage of healthy plants at the 28-day post-seeding in Fon race 2-inoculated soil. GBS-SNP data from genotyping-by-sequencing (GBS) for 205 F₂:₃ and 204 RIL families were used, and parental line genome sequences were used as references. Six GS models, including parametric (G-BLUP, BayesB, Bayes_LASSO) and non-parametric (Random Forest, SVM Linear, SVM Radial) methods, were tested. G-BLUP and Random Forest outperformed the other models, with correlations of 0.48 in the F_2:3 and 0.68 in the RIL populations, highlighting the GP efficacy in early-stage breeding for improving Fon race 2-resistance in elite watermelon cultivars.