AI-guided discovery of plant resistance genes

Plant diseases pose a clear and present threat to global food security. While durable genetic resistance is the most sustainable management approach, traditional methods to identify and deploy resistance genes (R-genes) remain slow, labour-intensive, and frequently outpaced by rapidly evolving pathogens. Thousands of valuable, unidentified R-genes lie within the germplasm of crop plants and their wild relatives. Our project exploits AlphaFold 3 (AF3) at an unprecedented scale to predict novel interactions between major crop pathogen effectors and a specific class of plant immune receptors called nucleotide-binding leucine-rich repeats with integrated domains (NLR-IDs). These receptors carry non-canonical domains that act as molecular decoys for effectors. Our goal is to dramatically accelerate the identification of functionally relevant NLR–effector pairs.

We assembled a curated dataset comprising thousands of NLR-ID sequences from hundreds of plant species and hundreds of effector sequences from major pathogens, including oomycetes, nematodes, aphids, and viruses. Using this dataset, we executed approximately 750,000 pairwise AF3 structural predictions, saving months, if not years, of traditional experimental screening. Interface confidence scoring was used to prioritize high-confidence candidate pairs, which we are currently validating through a combination of in vitro and in planta protein-protein interaction assays, alongside cell death and pathogenicity assays. Ultimately, our work provides a scalable, AI-driven pipeline for uncovering NLR–effector interactions. This enables the rapid development of disease-resistant crop varieties, reduces reliance on chemical pesticides, and enhances global food security.