Imagine waiting years for a medicine that never comes, or hoping for a treatment for a rare disease that almost no one is working on because the math simply doesn’t add up. That is the reality of drug discovery today. It is slow, staggeringly expensive, and prone to failure at almost every step. A molecule that looks perfect in a petri dish can turn out to be useless or even harmful in the human body. For decades, scientists have relied on trial and error, testing thousands of compounds in the hope that a handful might show promise. It is a process that can take more than a decade and cost billions of dollars, with no guarantee of success. Now a new British startup believes that artificial intelligence can change all of that. Potential Sciences, which officially launched on November 5, wants to use AI to speed up the slow and temperamental process of drug design. The company’s founder and CEO, Hatem Helal, is not promising a magic wand. Instead, he is betting that AI, trained on vast amounts of physical and chemical data, can help researchers predict which molecules are worth pursuing and which are doomed to fail long before they ever reach a laboratory. The goal is not just to make drug discovery faster, but to make it smarter, cheaper, and more humane—so that patients with rare conditions are not left behind simply because their diseases are not profitable enough to attract traditional pharmaceutical research.
The core idea behind Potential Sciences is deceptively simple: use the fundamental laws of physics to simulate how atoms and molecules interact, then use that information to train AI systems that can predict how a drug candidate will behave in the human body. Helal and his team plan to build large-scale chemical datasets by simulating interactions at the atomic level, then gradually moving up to small molecules, then to complex amino acids and proteins. The hope is that these physics-based simulations will generate enough high-quality data to teach AI models how to identify highly targeted drugs for specific diseases. If the system works, it could bypass many of the costly and time-consuming experiments that currently slow down drug development. Helal puts it bluntly: “We want molecules that are never going to work in the clinic to fail fast.” In other words, the company wants to weed out bad candidates early, before they waste years of research and millions of dollars. It is a bit like dating, but for chemistry. Instead of taking every potential match out for a long, expensive dinner, you first run a quick background check based on the laws of physics. If the molecule is never going to get along with human biology, you want to know that before you commit. The challenge, of course, is that human biology is enormously complex, and simulating it accurately is no small feat. But Potential Sciences is betting that with enough data and the right algorithms, AI can become a reliable matchmaker between molecules and diseases.
The startup has already attracted serious financial backing. It says it has seed funding in the tens of millions of dollars from SovAI, the UK government’s AI venture capital fund, as well as Twin Path Ventures and other investors. The founders declined to disclose the exact amount or the company’s valuation, but the involvement of a government-backed fund signals that this is not just a niche tech experiment. It is part of a broader national ambition to make Britain a leader in AI-driven science. For Helal, the potential impact goes far beyond faster drug development. By reducing the cost of experimentation, he argues, the company’s approach could make it economically viable for pharmaceutical companies to pursue treatments for rare diseases that have long been ignored. “We want to make it possible to have more bespoke treatments for diseases—this idea that medicines can be one-off molecular combinations that are tailored to your unique human biology,” he says. That vision is deeply personal for many patients and families. Rare diseases are often called “orphan” diseases precisely because they are abandoned by the pharmaceutical industry. There is simply not enough financial incentive to develop treatments for a condition that affects only a few thousand people. But if AI can lower the cost of discovery, the economics change. Suddenly, a treatment that would have been unprofitable might become viable. And in the longer term, Helal imagines a world where medicine is not mass-produced but personalized—where a drug is designed for your specific genetic makeup, your specific proteins, your specific disease. It is a beautiful vision, but it is also an audacious one.
The scientific community is watching with a mixture of excitement and caution. The central question is whether Potential Sciences can actually predict how well drug candidates will bind to disease-related proteins with the accuracy it hopes. That would require a marked leap forward from where the field stands today. Peter Wolynes, a professor of chemistry at Rice University, is cautiously skeptical. “I’m sure it will be an improvement,” he says. But currently, “the engineering is at the level of the Romans, not of the Renaissance, in terms of how protein engineering is done.” It is a vivid analogy. The Romans were brilliant builders, but they used brute force and empirical rules. The Renaissance brought a deeper understanding of the underlying principles, allowing for more elegant and precise construction. Wolynes is suggesting that even with AI, we are still in the early, somewhat clumsy phase of protein engineering. We can build things that work, but we do not yet fully understand why they work, and we certainly cannot predict everything with confidence. That does not mean the effort is doomed. It means that the road from promising idea to reliable technology is likely to be long and full of setbacks. The founders of Potential Sciences seem to understand this. They are not claiming to have solved drug discovery overnight. They are building infrastructure—datasets, models, and prediction systems—that they hope will improve over time. But the gap between a promising approach and a proven one is enormous, and many startups have stumbled in that gap before.
Potential Sciences is entering a broader movement that has been building for years. Industry luminaries like Demis Hassabis, who won the Nobel Prize for Chemistry for his work at Google DeepMind, have long argued that advances in AI will eventually lead to momentous scientific breakthroughs. AI has already transformed mathematics, helping researchers discover patterns and proofs that had eluded humans for decades. The hope among startups like Potential Sciences is that drug discovery will be one of the first scientific frontiers to be transformed by the technology. The logic is compelling. Drug discovery is essentially a search problem: you are looking for a molecule that fits a particular biological target, like a key fitting a lock. The number of possible molecules is astronomically large, far more than any human or laboratory could ever test. AI excels at searching vast spaces and finding patterns. If you can train an AI to understand the physics of molecular interactions, it could theoretically narrow down the search space dramatically, pointing researchers toward the most promising candidates. That is the dream. But the reality is more complicated. Biological systems are not static locks and keys. They are dynamic, messy, and constantly changing. A molecule that binds to a protein in a simulation might behave completely differently in the chaotic environment of a living cell. And the data used to train AI models is only as good as the physics and chemistry that generate it. If the simulations are incomplete or inaccurate, the predictions will be too. Still, the momentum is undeniable. Governments, investors, and researchers are pouring resources into AI-driven science, and the pace of progress is accelerating.
In the end, the story of Potential Sciences is a story about hope—hope that technology can make medicine faster, cheaper, and more personal. It is also a story about humility, because the scientific challenges are immense and the history of drug discovery is littered with failed predictions. The company’s approach, using physics-based simulations to generate training data for AI, is a promising middle path between pure theory and blind experimentation. It does not promise miracles. It promises better odds. It promises that some of the most painful failures in drug development can be avoided, and that diseases which have been neglected for too long might finally get the attention they deserve. Whether the company succeeds will depend on factors that are hard to predict: the quality of its data, the power of its algorithms, and the ability of its team to learn from mistakes. But even if Potential Sciences only partially succeeds, it could help shift the pharmaceutical industry toward a more intelligent, data-driven approach. And for the millions of people waiting for treatments that do not yet exist, that is a reason to pay attention. The road to a new medicine is long, but with AI, it may finally become a little shorter. The Romans built aqueducts that still stand today, but they did it with trial and error. The Renaissance built cathedrals with a deeper understanding of structure and design. Potential Sciences is hoping to bring drug discovery into its Renaissance—not by abandoning the lessons of the past, but by using AI to understand them more deeply. Whether that dream becomes reality remains to be seen, but the effort itself is a testament to the enduring human desire to heal, to create, and to push the boundaries of what is possible.
