Objectives: Develop DL architectures integrating text-mined, cheminformatic and biophysical descriptors to predict interface binding, allosteric signatures and resistance liabilities; jointly learn across peptide and small-molecule modalities; generate hybrid peptidomimetic design hypotheses; deliver consortium-wide prioritisation dashboards for hit triage and analogue selection.
Expected results: A resistance-aware AI workflow for non-ATP interface ligands, accompanied by public models/datasets and interpretable feature maps; prospective design case studies demonstrating improved optimisation trajectories in WP3 campaigns.
Secondments: VIB: Structural-class extraction and conformational-feature mapping from cryo-EM datasets (M20–23; 4 months); Skylight: SHG profiling of ligand-induced conformational changes (M31–33; 3 months).
Integration: Provides resistance- and allosterism-aware prioritisation engines for WP3 and closes the computational–mechanistic feedback loop with WP2.
Interdisciplinarity: Bridges AI/ML, cheminformatics, structural biology (cryo-EM constraints) and chemical-probe/screening environments.
