Methodology development

We build the tools we need to do the science. Better tools make work tractable that would otherwise stall: more researchers can run the same methods, larger problem sizes become feasible, and infrastructure stops being the rate-limiting step for discovery and engineering.

Open all-atom protein-structure prediction

Predicting how an immune receptor binds an effector is a core problem in the lab. AlphaFold 3 changed what is possible but its server access is restricted for commercial use. We package open all-atom models — Protenix, IntelliFold, Chai-1, OpenFold3, Boltz-2 — into Google Colab notebooks with a unified CSV input format, calibration-based parallel GPU scheduling, and automatic Google Drive upload. The notebooks are listed in Resources.

NLR catalogues and annotation

Identifying NLRs across plant genomes was a recurring bottleneck. RefPlantNLR (Kourelis et al., PLOS Biology 2021) collected the field's experimentally validated NLRs into one curated reference; NLRtracker built a sequence-search tool around it. Both underpin most NLR phylogenetics done in plant immunology today.

Receptor engineering toolkits

The lab maintains modular Golden Gate-compatible vector toolkits for cloning NLR variants and for high-throughput receptor screening, used routinely in the bioengineering work. Tools for screening pipelines are released to the community as they mature.

Computational pipelines for protein–protein interactions

Once we have hundreds of structure predictions, the next problem is reading them. ipSAE_batch combines interaction-scoring metrics with AlphaBridge-inspired visualisation across all the structure-prediction backends we use; it lets us triage and rank predicted receptor-effector complexes systematically.