Revealing The Catalyst Within

Generating small molecule medicines for any protein and every patient

The unreachable master switches of human health

The most shocking statistic in drug development is not failure rates or costs. It is that 80% of the proteins linked to disease have no existing medicines that target them. The age of omics created a window directly into the inner working of biology, but we are trapped outside. We know what we want to change, but the proteins that matter most in human disease remain unreachable by modern medicine.

Why? Drug discovery relies on finding molecules that can tightly bind to proteins. Small molecule medicines historically required deep, highly structured protein pockets to bind. The most druggable targets are those that evolved deep pockets to perform their function, such as kinases or proteases. But most proteins don’t have these kind of grooves and handholds. They act instead through broader interactions spread across their surfaces. For critical sensors, regulators, and transcription factors, which act as drivers in an enormous number of diseases, there’s nowhere for traditional molecules to grab hold.

Drug discovery has labelled these proteins ​“undruggable,” content with focusing on the few easy-to-bind targets and leaving millions of patients behind.

‘Undruggable’ is a flawed concept

Covalent medicines, where molecules bind irreversibly to their protein targets, emerged as a solution to this grand challenge of ligandability. Rather than relying on the reversible interactions with a deep pocket to tightly bind a medicine, covalency promised binding through a single, permanent bond. Reality did not meet that promise; the mindset that smooth surfaces were ​“undruggable” prevailed. Early adopters only added covalency as an accessory to existing molecules that could already bind deep pockets, not as the basis of a disruptive new way to create medicines.

What if we could flip the script? What if instead of relying on first binding to a protein pocket before covalent bond formation, the covalent reaction came first?

This biochemical insight was inspired by nature. Biology catalyzes bonds through enzymes that create electronically optimized microenvironments, straining and steering molecules at the quantum level, moving atomic orbitals into alignment to push otherwise impossible reactions forward. Yet, the same amino acids present in an enzyme’s catalytic site exist on the protein surface; when viewed this way, the surfaces of proteins are not smooth and ​‘undruggable’, but rich in electronic features that, given the right molecular substrate, can catalyze the formation of a covalent bond. We call these Latent Catalytic sites, and our reaction first chemical strategy is discovering the perfect substrate match for every protein surface.

With our Reaction First approach, we can design safe, stable, and drug-like molecules that are inert inside the body until they are activated by the right protein catalyst to form potent, selective, irreversible bonds — even on surfaces that were previously labelled ​‘undruggable.’ With this insight, the entire proteome is within reach of small molecule medicines and the most valuable targets in drug discovery — transcription factors, protein-protein interaction interfaces, and adaptor proteins to name a few — are not undruggable surfaces, but undiscovered sites of latent catalysis waiting for the right substrate.

Learning new rules for medicinal chemistry

To tap into the power of Reaction First drug discovery, we purpose-built a platform that parallelizes drug discovery with data generation. By combining innovations across chemoproteomics, hardware, software, quantum chemistry, and machine intelligence, we produce data at unrivaled richness, depth, and precision. The potency of each molecule in our library is simultaneously measured, down to the specific amino acid location, against thousands of proteins in relevant context. Our dataset is revealing the proteome’s substrate preferences: how similar microenvironments on divergent proteins can catalyze the same reactions.

With this data, our generative AI is learning rules of bond formation that traditional medicinal chemistry has often ignored. While classical drug discovery focuses on the surface interactions of atoms, our platform looks at reactions at the level of electrons. With a proprietary and continually growing proteome-wide bond formation dataset, our AI is learning how small changes in orbital energies shift a surface from inert to catalytic, the relationship between molecular quantum properties and transition-state, and how subtle topology templates reactions. Combining our empirical measurements with high accuracy synthetic quantum chemical data is enabling our model to generate molecules for targets it has never directly seen, but whose catalytic profile mirrors those already bound. Our reaction first approach is more than a biochemical insight; it is a new perspective unlocking generative medicines, designed not just from the atom, but from the electron.

The ultimate measure of Expedition’s success will not be the size of our dataset or the elegance of the molecules we generate, but the number of patients who no longer have to suffer from ​“undruggable” diseases. When we founded our expedition, we set out beyond the boundary of medicine to rethink the limits that have kept millions from cures.