From computational design to working biology
Across plant immunity, synthetic immune systems, and neglected disease, we follow the same loop: design computationally, then test rigorously to see which choices actually hold up.
De novo integrated domains for plant NLR receptors
Our central project uses de novo binder design (RFdiffusion, ProteinMPNN) to build synthetic integrated domains for the Pikm-1 NLR chassis, targeting the pathogen effector FoSSP17. X-ray crystal structures confirm side-chain-level accuracy of the designed interface. A key finding from this work is that binding and signal relay behave as separable properties of an integrated domain — a receptor can bind its target without necessarily triggering downstream immune signalling, and vice versa.
This work is described in a preprint (Xi, Bucknell, Watson et al., 2026), currently under review at PNAS. See Publications for the full reference.
RFdiffusion ProteinMPNN Pikm-1 chassis X-ray crystallographySUSS effectors and surface frustration
With Gregory Knight and Jonathan Heddle, we reviewed SUSS effector families and the concept of surface frustration as a lens for identifying broad-spectrum resistance targets — regions of an effector surface that are structurally strained and may be harder for a pathogen to mutate away from recognition.
Focused Review, The Plant Journal →Moving towards synthetic immune systems
A longer-term effort to rebuild plant immune signalling from modular, redesigned parts, rather than only adapting what evolution has already built. Details are under wraps while the work is ongoing — watch this space, or see Join Us, for upcoming PhD and postdoc openings tied to this project.
Synthetic immune signalling domains
We're interested in immune signalling domains as modular building blocks for synthetic biology, and what it takes to make them function as designed parts rather than only in their native context.
Protein design for neglected disease and diagnostics
Alongside our plant immunity work, we're applying the same design-then-test approach to targets in neglected tropical disease and to protein-based diagnostics. More to follow as this work matures.
Design, then test
Proteins are generated computationally, with the latest protein design and prediction softwares, on our BioDroid HPC system at the CPBM. Designs are then validated functionally in model biological systems, and validated through structural biology, biophysics and biochemistry. We treat the distance between a plausible AI-generated sequence and a functional validation as the central bottleneck the lab works on, rather than an assumption we can design around.