The $75M bet on antibodies the human immune system already made

$75 million. That is what it costs to take two antibody drugs into the clinic against the biggest drug in immunology.

Infinimmune, a startup out of Alameda, California, raised that Series A on Tuesday. The money funds two atopic dermatitis candidates, IFX-101 and IFX-201, into phase 1 testing in 2027. Both were found through a discovery engine that starts with human memory B cells and ends with a protein language model called GLIMPSE-1.

Atopic dermatitis is a crowded market by any measure. The biggest drug is Sanofi and Regeneron's Dupixent, an antibody that blocks the IL-4 and IL-13 pathways. Eli Lilly's Ebglyss, approved in 2024, also targets IL-13, and recent entrants such as AbbVie's Apogee are building long-acting IL-13 blockers. Dosing for the injectables ranges from every two to four weeks. The space is large and the standard of care is far from settled.

Its pitch is that the discovery method matters as much as the target. Screen the human immune system instead of engineering antibodies in mice, and the resulting drugs carry properties that make them faster to build and harder for patients to escape. The financing puts a number on that claim: $75 million from backers that include the company that owns the current standard of care in this disease.

The standard way to make a monoclonal antibody starts with mice engineered to carry pieces of the human immune system. Inject an antigen, wait for the mouse to raise an immune response, then humanize the result. It works. Almost every approved antibody came from this route. But the engineering happens on a framework that was never human.

Infinimmune starts somewhere else. Its Anthrobody platform screens millions of memory B cells, the immune cells that survive for decades after an infection and remember the antigens they once met. This living memory is what lets the body produce an antibody within days of re-exposure. Infinimmune captures those cells, sequences their antibodies and screens them against hundreds of targets at once. The result is a library of fully human antibodies selected by the human immune system itself, not reconstructed by a mouse.

The payoff, according to the company, is speed and drug quality. Antibodies pulled from native human repertoires come with the folding, stability and immune-compatibility patterns that human physiology actually tolerates. A drug that starts human does not need a humanization step that can shave potency or introduce developability problems.

That is the story in the press kit. The part the clinic will test is whether the screening data actually produces better drugs.

On the screening side, its single-cell sequencing platform measures millions of antibody sequences against hundreds of targets in a single pass, which is a step up from the older approach of immortalizing B cells into hybridoma cultures one at a time. The technology comes from a founding team with 10x Genomics roots, and it shows: the company reads the immune system the way 10x read single cells.

McDonnell is explicit about the ambition. The company has said it is not a platform company, it is a therapeutics company with a productive platform. That distinction matters for investors, because platform stories have a way of underdelivering on the medicine. Infinimmune is betting the pipeline, not the process.

The second half of the engine is GLIMPSE-1, a protein language model the company launched in June 2025. Unlike general-purpose protein models trained on databases of everything from bacteria to viruses, GLIMPSE-1 was trained solely on paired heavy and light chains from human memory B cells. The pair matters: standard design approaches treat the heavy and light chains separately, but the way they interact determines stability, developability and binding. A model that sees the pair together captures what a chain-by-chain approach misses.

In a bioRxiv preprint from June 2025, the company reported three results. First, GLIMPSE-1 scored best-in-class on antibody humanization benchmarks against nine leading antibody language models. Second, it produced affinity-optimized variants in the sub-nanomolar range while preserving developability. Third, it generated functional variants with under 90% sequence identity to their parents, which is a way of saying the model can propose genuinely new antibody shapes, not just tweak existing ones.

The most striking claim is the data efficiency one. Using 95% less training data than the prior state of the art, the model matched the humanization performance of Sapiens, Merck's best-in-class antibody model. If that holds up, it points to a real advantage: Infinimmune's proprietary human-repertoire data is the moat, and the model is the tool that spends it efficiently.

None of this is clinical validation. GLIMPSE-1 has engineered the startup's first pipeline assets, and those assets are the ones now heading to phase 1. The model is a discovery accelerator, not a proof of efficacy. What the preprint establishes is a ceiling on how well the platform can design candidates. What the clinic establishes is whether those candidates work in people.

The Series A carries two lead programs. IFX-201 targets IL-13, the same cytokine that anchors Dupixent's mechanism. On paper this looks like a me-too play against a blockbuster. McDonnell frames it differently: many patients stop responding to one IL-13 blocker and then respond to a different one that hits the same target. The clinical history of eczema shows that switching mechanisms within a target class works. A second-generation IL-13 antibody with better dosing or durability has a real opening.

IFX-101 is the more interesting bet. It blocks IL-22, a cytokine that drives the barrier breach, the breakdown of the outer skin layer that defines severe atopic dermatitis. There is no FDA-approved therapy for IL-22 today. The company calls IFX-101 a potential first and best-in-class therapy, and that claim is what the market should interrogate hardest.

The logic is coherent. IL-13 drives the inflammatory flare, while IL-22 drives the skin-barrier defect. Two different pathologies, two different targets, one disease. Dupixent and its peers fight the inflammation but leave the barrier repair to the patient's biology. IFX-101 addresses the barrier itself. Patients who do not respond to IL-13 therapies, McDonnell argues, often have inflammation driven by IL-22. That is a testable hypothesis, and phase 1 will begin testing it next year.

The platform has produced more than the two lead assets. Earlier-stage antibodies target APRIL, a signaling protein implicated in IgA nephropathy, and IL-17F for ulcerative colitis. One more asset remains undisclosed. The pair of lead programs is enough to justify the round; the rest is optionality.

The market it is entering already has a shape. Dupixent is the blockbuster, approved first and still the default prescription for moderate to severe atopic dermatitis. Ebglyss and Nemluvio reached the market later, both built around IL-13 blockade, and both have been priced as follow-ons rather than replacements. Adbry rounds out the group as an IL-13 option with a different dosing profile. The common thread is dosing: every approved injectable requires a shot every two to eight weeks, for a chronic condition that lasts years.

That cadence is the opening. A drug that holds patients with less frequent injections changes the economics of adherence, and adherence, not headline efficacy, is where the real-world value of eczema drugs gets lost. Patients who miss doses flare; flares lead to systemic therapy or discontinuation. A long-acting IL-13 antibody has been the quiet favorite to capture that dynamic, which is why AbbVie paid up for Apogee, whose lead program is a long-acting IL-13 blocker with phase 2 data supporting dosing every three to six months.

It sits between two games. IFX-201 enters the IL-13 segment where Dupixent, Ebglyss, Nemluvio and Apogee already compete, betting that a fully human antibody with faster discovery can win on durability rather than raw efficacy. IFX-101 plays a segment with no approved players at all, betting that IL-22 blockade addresses the barrier-breakdown pathology the current drugs leave completely untreated. Two targets, two different market positions, one platform funding both.

The financing has a wrinkle that separates it from the usual biotech round. Regeneron Ventures co-led the $75 million alongside Playground Global. Regeneron's parent company makes Dupixent, the eczema blockbuster whose mechanism Infinimmune is trying to improve on. A company financing the development of a rival to its own flagship drug is unusual. It is also rational.

Inflammation is a portfolio business. Regeneron knows IL-13 biology better than anyone, and it knows which patients the current crop of drugs fails. Backing a second-generation antibody is a hedge: if IL-13 blockade gets displaced, Regeneron holds a position in the replacement. The venture arm also gets deal flow and insight into where the immunology field is heading. The round is a bet on the platform as much as on any single asset.

The rest of the syndicate reads like a who's who of biotech later-stage capital. RA Capital Management joined as a newcomer. Existing investors returned, including Merck & Co., which is also a partner. Merck Global Health Innovation Fund, Everbright Biofund, Civilization Ventures and Pear VC all participated.

The Merck connection is the second validation. It has signed antibody discovery pacts with Merck & Co. and, earlier, with Immunome to find new antibody drugs for unnamed immune conditions. These deals do not just bring money. They bring external validation that the platform works on targets chosen by other companies, and they put the model in front of pharma's best antibody engineers.

The company's earlier rounds point the same direction. Playground Global led a $12 million seed round in December 2022, back when the company was five technologists with 10x Genomics origins. Eighteen months later it is a $75 million Series A with a Big Pharma partner and a Regeneron co-lead. The trajectory is the pitch.

The honest version of this story starts with what can go wrong. Atopic dermatitis has been crowded for a decade, and the entrants keep coming. A biotech named Apogee, recently acquired by AbbVie, is developing a long-acting IL-13 blocker with phase 2 data supporting dosing every three to six months. If the market's next shift is toward infrequent dosing, IFX-201 enters a race where someone already holds a lead.

IFX-101's first-in-class positioning cuts both ways. Being first to an undrugged target means no validation exists. IL-22 blockade could fail the way dozens of undrugged targets have failed before: the biology is compelling in the dish and the mouse, and the human readout shows nothing. The company's own caution reflects this. Both antibodies could enter the clinic as soon as 2027, but the language is deliberately hedged.

There is also the financing runway question. $75 million at current biotech burn rates funds discovery, manufacturing and early clinical work, but a pair of phase 1 studies in a competitive market rarely comes cheap. The CEO has said the cash lasts through meaningful value inflection points, which is a careful way of saying the money gets the company to data, not to approval. Another raise will be needed, and the terms of that raise will depend on what the 2027 readouts show.

We are going after a pair of targets that play different roles in that core disease biology. We are looking at this as an opportunity to develop complementary molecules for a disease that is really complex, that is very multifactorial.— Wyatt McDonnell, CEO, Infinimmune

What the round actually buys is a test of the platform thesis. GLIMPSE-1 trained on 95% less data than Merck's Sapiens and matched it. Anthrobody screens millions of memory B cells where older methods screened dozens. Both claims are promising and both are unproven in humans. The 2027 phase 1 data is the first real checkpoint, and it will be public and verifiable.

The atopic dermatitis drug market rewards durability, not just efficacy. The current generation of antibodies works, but it asks patients to inject themselves every two to eight weeks for life. The next generation competes on how long the effect lasts and how many patients keep responding. Infinimmune's bet is that human-derived antibodies with an LLM that designs them faster will win that race. The $75 million says the market considers the bet fundable.

Regeneron's participation is the detail that should keep investors interested. Big Pharma venture arms do not usually back challengers to their own blockbusters unless they believe the challenger has a real shot. The co-lead is the strongest independent signal in the entire round, stronger than the valuation or the syndicate. It means the current champion thinks the challenger's platform might be the next platform, and that signal deserves attention.

Sources