We are building digital twins of disease, starting with highly accurate models of protein behavior and drug binding inside cells. Our approach is built on a combination of thorough data science, molecular biology, and artificial intelligence. Combined, these technologies form a platform for high throughput discovery of safe, effective therapeutics.
We've created a knowledge graph, TOME, that integrates disparate information on drugs, targets, and their interactions. TOME connects the chemical, functional, and structural properties of targets to how they interact with drugs. TOME reveals how variability in a target can heavily impact drug binding, with some drugs only becoming active after a target changes, whilst other drugs are completely nullified.
Importantly, this target variability is not accounted for in any computational models that predict drug-target interactions because we lack sufficient data on target variability in the context of drug-target binding.
We are employing experimental molecular biology to fill this gap in TOME, generating 100K data points on drug-target binding where target variability is accounted for. This is a world-first, proprietary data set that will allow for powerful, unique models for drug-target interaction prediction. As we grow, our data generation will increase to millions of data points, allowing us to create increasingly powerful AI.
With data on drug-target binding that accounts for target variability, we can create powerful models that predict drug binding against variable targets. These models leverage deep graph learning to accurately represent the features of drugs and targets. In this way, we are creating high fidelity models of targets that more realistically represent how they behave inside the body. We are creating digital twins of targets. From these models, we can scour the drug universe for drugs that bind all variations of the target. As these binding models are more specific, we can discover safer, more effective therapeutics.
Modelling target variability in the context of drug-target binding is only the start. As we scale we can continuously add layers of information about drug-disease interactions to create ever more powerful models for drug discovery. At every phase, our focus will be high fidelity data to create accurate digital twins of disease.