Our ideas about diseases outpace our ability to drug them. 80% of protein targets are currently untouched because they don’t fit conventional ideas about what proteins are amenable to small molecule medicines. For these proteins, it’s hard to find a chemical starting point, and for any protein, it takes years of effort, a few dozen compounds at a time to turn those starting points into medicines.

We combine machine learning with massively parallel biochemical tools such as DNA Encoded Libraries (DELs) and Affinity Selected Mass Spectrometry (ASMS) to analyze more compounds more efficiently than ever before. By working with large datasets throughout our search process and letting our machine learning model guide our experiments, we are able to find molecules for the hardest problems in drug discovery, bringing first-in-class and best-in-class treatments to patients.

Recent advances in functional and genetic screening powered by innovations such as CRISPR have identified a tantalizing number of proteins that could meaningfully alter disease progression. Still, new medicines, particularly the small-molecule medicines that account for the majority of FDA approvals, take a decade or more to develop. At Anagenex, we integrate machine learning, biology, and chemistry to address historically intractable medical problems and bring medicines to patients faster.

First, we test billions of custom-synthesized compounds to see which ones are likely to modulate a protein “target” we believe to be important in some diseases. For any target, we run dozens of experiments at a billion compound scale, generating rich, high-quality datasets. Those datasets train proprietary neural networks to understand what compounds are promising. Finally, we use those machine learning models to design new multimillion compound experiments, synthesizing and then testing those compounds to repeat the cycle. Every iteration improves and bring