Machine learning empowers precise discovery of disease-resistance genes in plants

Abstract

Identifying plant disease-resistance genes is essential for understanding the plant immune system and accelerating the breeding of disease-resistant crops. There is a pressing need for a method capable of accurately identifying plant disease-resistance genes on a genome-wide scale. In this study, we propose evolutionary scale modeling for LRR (ESM-LRR), a deep protein language model designed to accurately predict LRR domains, which are substantially variable structures in disease-resistance proteins. ESM-LRR achieved its highest F1 score of 0.80 on a test set using 90% identity as the matching threshold. Building on ESM-LRR, we developed R-Predictor, a plant disease-resistance gene predictor to simultaneously annotate 15 diverse domain topologies, covering characterized resistance genes across the whole genome. R-Predictor integrates 4 modules, each employing superior methods that outperform existing methods (achieving F1 scores of 0.89 for RLKs and 0.88 for NLRs), demonstrating its high accuracy and practicality in annotating plant disease-resistance genes. R-Predictor integrated with gene expression profiles to identify candidate R genes associated with grape gray mold and downy mildew, outperforming existing methods and detecting dozens of candidate R genes. Overall, this study presents a novel approach to advancing our understanding of plant immunity and facilitating crop breeding for disease resistance.