Quantum chemistry, AI, and a stubborn belief that “undruggable” is just a target you haven’t watched move yet.
Somewhere on a server in New York, a protein is doing the one thing every textbook diagram forgets to mention: it is moving. Folding, breathing, opening a pocket that exists for a fraction of a fraction of a second and then closing it again. Most of drug discovery never sees that pocket. Fathom Therapeutics built a machine that does almost nothing else.
The company spent its early life as Atommap, a name that suggested cartography more than ambition. In April 2026 it shed the old skin, raised $47 million, and re-emerged as Fathom - a word that means both “to measure the depth of” and “to finally understand.” Both readings are on the nose, which is unusual for a rebrand and forgivable for a company that designs molecules atom by atom.
What it actually does is less poetic and more useful: it simulates how proteins behave inside living cells, then designs small molecules, degraders, and molecular glues to act on the targets everyone else filed under “impossible.” That is the whole company, really. Watch the motion. Find the gap. Make the drug.
For decades, drug design leaned on the crystal structure: a protein frozen, photographed, and treated as if that single pose were the truth. It is a convenient fiction. Real proteins flex and shift thousands of times a second, and a great many of the most important disease targets - mutant oncogenic proteins, the slippery surfaces where two proteins meet - look hopeless precisely because their useful pockets only appear when they move.
The industry's answer was usually brute force: screen millions of compounds and pray. Expensive, slow, and oddly faith-based for a science. The other answer, classical physics simulation, was accurate but glacial - the kind of calculation that finishes around the time the patent expires.
So the field was stuck between fast-and-blind and accurate-and-late. Fathom's founders had spent their careers on both sides of that wall - and decided the wall was the product.
Fathom was founded by three computational scientists who are, collectively, slightly intimidating. Huafeng Xu, the CEO, helped pioneer the Anton supercomputer - a machine built for the sole purpose of simulating molecular motion. Yujie Wu, the CTO, came through Schrödinger and Roivant Discovery. Jesus Izaguirre, Chief Computational Scientist, arrived from Silicon Therapeutics and Roivant. Between them and the team they assembled, they have pushed 19 drugs into the clinic. Seven of those have FDA approval - a record most platform companies can only describe in the future tense.
Helped pioneer the Anton supercomputer for molecular dynamics. The person most likely to tell you a protein is “just getting started.”
Built systems at Schrödinger and Roivant Discovery before betting on motion over snapshots.
From Silicon Therapeutics and Roivant. Turns equations of molecular movement into things you can actually run.
The platform is called Microcosmos, and Xu describes it, without much modesty, as “a world model of drugs in living cells.” It fuses quantum chemistry with AI to simulate protein motion at atomic resolution - and, crucially, reports doing it up to 10,000 times faster than conventional approaches without trading away accuracy. That speed is the whole point. It turns “too slow to bother with” into “run it before lunch.”
Because it models movement rather than a single pose, Microcosmos surfaces the transient binding sites that static models never see - the pockets that flicker open for an instant. Then it switches from observation to invention, using generative design to propose molecules built for those fleeting opportunities. The whole thing runs as a lab-in-the-loop: the computer predicts, the wet lab tests, the results feed back, and the next cycle is smarter than the last.
Simulates proteins flexing in real time at atomic resolution - up to 10,000x faster than classical methods.
Reveals transient pockets and protein-protein interfaces invisible to static structure models.
Generative design proposes small molecules, degraders, and molecular glues built for those pockets.
Lab-in-the-loop cycles pit prediction against wet-lab reality and learn from the gap.
A computational drug-discovery company sets out to map molecules atom by atom - the quiet, technical groundwork that the rebrand would later stand on.
The platform reportedly designs potent degraders for a previously “undruggable” target in roughly six weeks - the kind of timeline that makes investors put down their coffee.
Atommap becomes Fathom Therapeutics, trading a map for a depth gauge and signalling an appetite for the problems labelled impossible.
Sutter Hill Ventures leads, with Chemistry, Alexandria Venture Investments, and Empire State Development’s NY Ventures joining. Mandana Honu, PhD named Chief Business Officer.
The first internal program advances into animal efficacy studies as the company scales its lab-in-the-loop and discovery partnerships.
Platforms are easy to pitch and hard to believe. Fathom's case rests on three things you can count: a track record, a speed claim, and a fresh pile of capital that says serious people checked the math.
The backers tell their own story. Sutter Hill Ventures rarely chases trends; Alexandria builds the literal real estate of biotech; New York's NY Ventures planted a flag for the city's life-science ambitions. An oversubscribed round in a cautious market is the market's way of saying it would like to buy more, please.
Strip away the physics and the mission is plain: design medicines for diseases that have resisted them, by understanding molecules well enough to predict what they will do in a living cell before a single dose is made. The undruggable list - mutant oncogenic proteins, protein-protein interfaces, the targets that have humbled the industry - is exactly where Fathom wants to spend its time.
It is a healthcare company wearing a software company's confidence, which is either the future of drug discovery or a very well-funded hypothesis. Fathom would argue the difference is just a few more turns of the loop.
Return to that protein on the server, still mid-motion, still opening and closing a pocket the textbooks ignore. The difference now is that someone is not only watching - they are designing a molecule to slip in during the open moment, and they can do it before the protein finishes the gesture.
If Microcosmos works at the scale Fathom promises, the consequence is not a better screening assay. It is a shift in what counts as a target at all. Diseases written off as chemically unreachable get a second look. The list of “undruggable” proteins gets shorter every time the loop turns.
For now, Fathom Therapeutics is a 22-person company with a big claim, a real track record, and $47 million to find out if the claim holds. The protein is still moving. This time, that is the whole opportunity - and Fathom is the one paying attention to the gaps.