The Pocket that Wasn’t There
The last time a drug was discovered by staring at a protein and guessing where a molecule might fit, the process took fifteen years and cost more than a billion dollars. That was the old way. Before the machines learned to see the invisible.
For decades, drug discovery was a game of hunting in the dark. You had a protein that caused disease. You had a library of molecules that might bind to it. You spent years testing each one, hoping to find a key that fit a lock you could barely see. The problem was that many locks had hidden chambers — pockets that only appeared when the right key was already inside. You could not find what you could not see.
That is the threat. The human eye, even aided by the best microscopes and crystallography, can only see what is there. But proteins are not static. They breathe, they shift, they fold and unfold. Some pockets do not exist until a molecule coaxes them open. These are the cryptic pockets, and they have been the graveyard of countless drug programs. Scientists knew the disease target. They knew the protein. But the lock was invisible.
The mechanism that changed everything came from a lab in London that had already won a Nobel Prize for solving a different problem. AlphaFold had cracked protein folding — predicting the three-dimensional shape of a protein from its amino acid sequence. It was a triumph. But as Adrian Stecuła, a group leader at Isomorphic Labs, explained, proteins do not exist in a vacuum. [1] They interact with other molecules, with nucleic acids, with small molecule ligands, with ions, with other proteins. AlphaFold could tell you the shape of the protein alone. It could not tell you where a drug might fit, or how tightly, or whether the fit would change the protein’s behavior.
The Isomorphic Drug Design Engine, or IsoDDE, was built to go further. It is not just a structure predictor. It is a unified system that predicts three things: where the pocket is, how the ligand binds, and how strong that binding will be. But the real test came when the system was asked to find something that had never been seen before.
A technical report published in February described the IsoDDE system, which was applied to a protein called cereblon. [1] Cereblon is a well-studied protein involved in the degradation of other proteins — a pathway that drugs can hijack to destroy disease-causing proteins. The new pocket had never been observed. No one knew it existed. The system identified a previously unknown binding pocket on cereblon, a protein central to targeted protein degradation therapies It found the invisible lock.
This is where the adaptation begins. The old model of drug discovery assumed that druggable targets were those with obvious pockets. If a protein had no visible cavity, it was considered “undruggable.” But the human genome contains thousands of proteins associated with disease, and many of them have no obvious binding site. IsoDDE changes that calculation. It can find pockets that are not there until they are needed. It can predict how a ligand will bind to a pocket that only opens in the presence of that exact ligand. It is a system that sees the future of the interaction before the interaction happens.
The promise is not just faster drug discovery. It is the possibility of targeting diseases that have been out of reach for decades. Many diseases have known associated proteins — proteins that, if modulated, would help patients. But those proteins lack the kind of pocket that traditional drug design can exploit. IsoDDE opens the door to those proteins. It makes the undruggable druggable.
But the hope is tempered by a clear-eyed understanding of what the technology can and cannot do. Stecuła is careful to note that structure prediction alone does not solve drug discovery. [1] The system must predict not just where a molecule binds, but how it binds, how tightly, and what happens when it binds to other proteins in the body. There are many endpoints that matter, and IsoDDE is only beginning to address them.
The next scientific question this discovery makes possible is whether the same approach can be applied to other types of therapeutic modalities. The same system that found the cryptic pocket on cereblon could one day design drugs that work by mechanisms we have not yet imagined, including interactions with antibodies, molecular glues, and peptides It can also predict interactions with antibodies, molecular glues, and peptides. The same system that found the cryptic pocket on cereblon could one day design a completely new class of drugs that work by mechanisms we have not yet imagined.
The door that just opened is not just a door to faster drug discovery. It is a door to a new way of thinking about proteins and their hidden possibilities. The invisible locks are now visible. The question is what keys we will design to fit them.