Truace
Subscribe

Machine Learning and Statistical Models for Genomic Selection of Phytophthora Fruit Rot Resistance in Watermelon

Breeding for Phytophthora fruit rot (PFR) resistance caused by the devastating pathogen Phytophthora capsici remains challenging in watermelon mainly due to phenotyping constraints. To overcome these bottlenecks, we developed and validated genomic selection models using an interspecific recombinant inbred line (RIL, F 11 ) population derived from USVL531-MDR (resistant, Citrullus mucosospermus ) × USVL677-PMS (susceptible, Citrullus lanatus ) and a segregating F 2 population (USVL531-MDR × 'Calhoun Grey'). Matur…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1299· cite & verify
Machine Learning and Statistical Models for Genomic Selection of Phytophthora Fruit Rot Resistance in Watermelon
Diseases of watermelons by Orton, W. A. Meier, F. C.. Public domain

In brief

By October 2026, researchers had developed and validated genomic selection models to address phenotyping bottlenecks in breeding watermelon for resistance to Phytophthora capsici fruit rot. Using an interspecific RIL population and an F2 population derived from resistant USVL531-MDR and susceptible parents, they phenotyped inoculated mature fruits and tested fourteen models including Bayesian, regression-based, RKHS, Random Forest, SVM, and XGBoost.

The work matters because it shows machine learning and statistical genomic prediction can substitute for difficult field phenotyping and accelerate selection of resistant watermelon lines, with high accuracy in RILs and up to 85% coincidence with phenotypic selection. Uncertainty remains about transferability, as accuracy dropped to moderate levels in the segregating F2 population.

Main points

  1. Study used interspecific RIL F11 population USVL531-MDR x USVL677-PMS and F2 population USVL531-MDR x 'Calhoun Grey' phenotyped for lesion diameter, pathogen growth, and sporulation.
  2. Fourteen models were compared: ten parametric Bayesian and regression-based, one semi-parametric RKHS, and three machine learning algorithms Random Forest, SVM, XGBoost.
  3. F2 population showed lower predictive performance than RIL, with best r = 0.476 for GBLUP in 10-fold and r = 0.433 for Bayesian ridge regression in 5-fold cross-validation.

The gain

Genomic selection models including machine learning algorithms predicted PFR resistance in watermelon, achieving high cross-validated accuracy in the RIL population and capturing top-performing lines.

The rundown

Researchers artificially inoculated mature fruits from an interspecific RIL F11 and an F2 cross involving resistant Citrullus mucosospermus USVL531-MDR and susceptible Citrullus lanatus lines, measuring lesion diameter, visible pathogen growth diameter, and sporulation intensity under controlled conditions.

Fourteen genomic selection models were evaluated with 10-fold and 5-fold cross-validation, and coincidence analysis assessed overlap between genomic and phenotypic selection at 10% and 20% intensities, establishing a framework for PFR resistance breeding.

Sources

  1. Peer-reviewedPhytopathology®2026-10-03

The debate