Diagonal Therapeutics

A handful of antibodies, one improbable instruction: turn it back on

Inside a lab in Watertown, Massachusetts, a screening run is sorting through billions of antibody combinations. It is not looking for the molecules that bind hardest, or block best. It is hunting for the rare few that grab two cell-surface receptors, pull them together, and restart a conversation the body stopped having. That is the entire premise of Diagonal Therapeutics - and for most of pharmaceutical history, it has been the road not taken.

By early 2026 the company has roughly $253 million in the bank, a lead drug called DIAG723, and a first-in-human trial on the calendar. Twenty-eight people. One mechanism almost nobody else is industrializing. The pitch is unfashionably simple: stop silencing disease, and start restoring the signal disease silenced.

The drug industry got very good at saying "no"

Antibodies are biotech's most reliable workhorse, and they are almost all antagonists. They neutralize a virus, mop up an excess protein, gum up a receptor so it cannot fire. Useful - and, it turns out, only half the story. A large class of severe diseases isn't caused by too much signaling. It's caused by too little. The pathway that should keep blood vessels orderly goes quiet, and the body has no built-in volume knob to turn it back up.

Take hereditary hemorrhagic telangiectasia (HHT), which affects more than 330,000 people across the U.S. and Europe. When ALK1 signaling in the cells lining blood vessels falters, fragile, malformed vessels form and bleed. Pulmonary arterial hypertension (PAH), which hits around 200,000 people worldwide, runs on a related fault in the same neighborhood. The standard playbook - block something - doesn't fit a disease whose problem is a signal that already went missing.

What if you could make antibodies that activate?

Agonist antibodies - the kind that switch signaling on instead of off - have always been the white whale of the field. They're real, they're rare, and they're maddeningly hard to find on purpose. Alex Lugovskoy had spent two decades watching that gap from the inside: COO at Dragonfly Therapeutics, CDO at Morphic Therapeutic, VP of Therapeutics at Merrimack, and earlier, drug discovery at Biogen. He didn't think the molecules were impossible. He thought the search was the problem.

With Atlas Venture, he founded Diagonal in 2022 around a single conviction: if activation depends on physically clustering receptors in just the right geometry, then finding those antibodies is a search problem at industrial scale - and a search problem can be engineered. The wordmark hides the joke. Most of the field runs straight at a target to block it; Diagonal cuts across the field at an angle to switch it on.

Clustering antibodies, made on an engine

Diagonal's platform has an unglamorous, telling name: the DIAGONAL Product Engine. It pairs computational design with wet-lab screening to sift billions of antibody combinations for the rare ones that bind two or more receptors and cluster them into an active configuration. Computation narrows the field; experiment confirms what actually fires. The output is a class the company calls clustering antibodies - multi-specific, agonist, and disease-modifying by design.

The first molecule out of the engine is DIAG723. It restores dysregulated ALK1 signaling - the exact fault behind both HHT and PAH - which is why one antibody can credibly aim at two diseases. In preclinical models the company reports it both prevented and reversed disease pathology, and restored normal signaling in cells taken from patients. The next, and far harder, test is people.

Two products, one idea: build the machine that finds the medicine, then let it find the medicine.

From stealth to clinic-bound in under two years

A two-year sprint that most biotechs spend just clearing their throat.

The money has opinions. So does the biology.

Conviction in biotech is measured in two currencies: data and dollars. On the dollar side, Diagonal's cap table reads like a who's-who that doesn't usually agree on the same idea. Atlas Venture and BVF Partners launched it. Eighteen months later, Sanofi Ventures and Janus Henderson co-led an oversubscribed Series B, joined by Deep Track, EcoR1, Logos, Balyasny and Woodline - crossover and strategic investors who tend to show up when a clinical story is about to get real.

On the data side, the claims are still preclinical - which is the honest place to be. DIAG723 has shown disease-modifying activity across multiple HHT and PAH models and restored signaling in patient-derived cells. That is a strong hand. It is not yet a winning one. The clinic decides.

One mechanism, many diseases - if it holds

The bet underneath the bet is repeatability. If clustering antibodies can fix ALK1 signaling in vascular disease, the same engine should reach other conditions where a pathway has gone quiet - the company points toward hematology, hepatology and nephrology. A platform is only worth its name if it produces a second molecule, and a third. DIAG723 is the proof of principle the whole thesis rests on.

It's worth being skeptical here, and the company seems to know it. Plenty of "platforms" turn out to be one good drug wearing a lab coat. Diagonal's answer is mechanistic rather than rhetorical: the way it finds agonists isn't specific to one disease, so the second target is a search-engine query, not a fresh act of luck. That claim gets tested the moment a second program enters the clinic.

Back to that screening run

Return to the lab in Watertown. The screen is still grinding through billions of combinations, still hunting for the rare antibody that pulls two receptors together and restarts a signal. Two years ago that was a research curiosity. Now it has a $253 million budget, a drug headed for first-in-human testing, and a roomful of investors waiting to see whether a quieted pathway can be turned back up in an actual patient.

If DIAG723 works, the scene changes meaning entirely. The thing being searched for stops being a thesis and becomes a category - a way to treat the large, stubborn class of diseases defined by signals that went missing. If it doesn't, Diagonal will have run one of the cleaner experiments on whether agonist antibodies can be industrialized at all. Either way, the question is no longer hypothetical. For the people living with HHT and PAH, that is the part that counts.