Decoding Biomolecular Interaction from Sequence Data

Life is a conversation. Inside every cell, biomolecules “talk” by binding to one another, a process that shapes health, disease, and how medicines work. Yet predicting these interactions has traditionally depended on costly 3D structural data and intensive computation. To break this bottleneck, I developed PSICHIC (pronounced “psychic”), a family of AI models that predicts biomolecular interactions directly from sequences—the simple letter codes of life. Like language models that learn patterns and solve problems from text, PSICHIC learns the underlying “grammar” of binding from abundant, low-cost sequence data, enabling rapid, scalable screening and prioritisation of candidates at the click of a button. By bypassing 3D structures, PSICHIC helps make the discovery of new treatments faster, cheaper, and more precise.