Drug discovery · Machine learning medicinal chemistry
PostEra
An AI chemistry company that gave its most famous work away for free. Seven years in, it has sold two preclinical fertility molecules to Merck KGaA, and not one molecule it designed has yet reached a human.
Hundreds of chemists, stuck at home
In the spring of 2020 the world’s medicinal chemists were sent home and their projects were paused. Alpha Lee noticed.
“We realised there were hundreds of chemists sitting at home, with their projects on hold,” he told Chemistry World that April. His company, then three people and about six months old, set up a non-profit project and asked those chemists to design molecules that would jam the main protease of SARS-CoV-2, an enzyme the virus needs to copy itself and one humans do not have.
By April 7 that year, 400 contributors had sent in more than 3,500 suggestions. PostEra’s software ranked them by how easily they could be made, wrote the synthetic routes, and shipped the list to Enamine in Kyiv. The first hundred compounds reached the University of Oxford on April 8. Sorting the first 2,000 submissions took about 48 hours; the team reckoned people would have needed three weeks.
That project became the COVID Moonshot. It is the reason most chemists have heard of PostEra, and the company never earned a dollar from it.
The gap it is aimed at
Most machine learning in drug discovery answers one question: which molecule should we make next? PostEra was built around the awkward follow-up question.
“Most AI approaches to drug discovery focus on answering the question of what compounds to make, but miss how to make those compounds,” Lee said in 2020. A model that proposes a molecule no chemist can build in a reasonable number of steps saves nobody any time. PostEra’s first product, a route-planning tool, was trained on roughly 9 million reactions pulled from US patents. In August 2019 the underlying algorithm was reported to beat eleven MIT chemists at predicting the outcome of 80 random reactions, 87.5 percent accuracy against 76.5 for the best of them.
Design and synthesis planned together rather than in sequence. The platform is now called Proton.
Two debating partners and a graduate student
Lee and Aaron Morris met reading mathematics at Oxford in the early 2010s, as partners in the debating club. “We were the only two science kids trying to do public speaking,” Morris told Chemistry World. Lee went to Cambridge and started a research group. Matthew Robinson joined it as a student. Morris went to Goldman Sachs.
In October 2019 the three incorporated PostEra in London to commercialise Lee’s Cambridge work. By early 2020 the UK shop was closed, a US entity was open, and they were in Y Combinator’s winter batch. The pandemic then stranded Morris and Lee in California for nearly six months, designing molecules from a Bay Area Airbnb.
Free work as a sales pitch
The Moonshot gave PostEra something an ordinary contract never would: a public, auditable record of its software working on a real program, with nothing withheld.
“Some of those really fortuitous events, like the open-science COVID Moonshot, like the NIH grant, we wouldn’t make a dollar of profit from any of that, but it gave us some of the best industry validation for our AI possible, because we had nothing to hide,” Morris told C&EN in 2025.
Pfizer called in late 2020. In January 2022 it widened the deal into a joint AI Lab, paying $13 million up front against up to $248 million in milestones, and PostEra closed a $24 million Series A led by the Clermont Group the same day. In May 2022 the NIH named Lee as one of three principal investigators on an antiviral drug discovery center. Amgen signed in November 2023 for up to five small molecule programs on undisclosed terms. In January 2025 Pfizer expanded again, to a headline $610 million, adding antibody-drug conjugate payloads and $12 million more up front.
Then the company turned inward. It moved into a Cambridge coworking space, stopped taking venture money, and pointed Proton at conditions that affect women: fertility, and what it now calls polyendocrine metabolic ovarian syndrome. On September 9, 2026, EMD Serono, the US and Canadian healthcare arm of Merck KGaA, bought two of those programs outright.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| COVID Moonshot output | Published in Science, Nov 10, 2023: more than 18,000 compound designs, more than 2,400 synthesized compounds, more than 490 ligand-bound x-ray structures and more than 10,000 biochemical measurements, released in real time with no intellectual property attached. Robinson, Morris and Lee are named authors. | Public record |
| Fate of the Moonshot lead | DNDI-6510, the frontrunner Mpro inhibitor, “ultimately failed in early testing,” per DNDi’s portfolio page. | Public record |
| Fate of the backup | ASAP-0017445 was nominated a pre-clinical candidate by DNDi in September 2025 after independent expert review; structure disclosed March 2025. It has not been in a human. Its backbone came from crowdsourced Moonshot designs and it was optimized by the ASAP consortium, in which PostEra is one of eight named partners. | Public record |
| NIH antiviral center | NIAID awarded about $577 million across nine AViDD centers in May 2022. ASAP received an initial $68 million for three years, with Chodera (MSKCC), Perry (DNDi) and Lee (PostEra) as principal investigators. Funded 2022 to 2025. | Public record |
| Peer-reviewed partner chemistry | A SARS-CoV-2 papain-like protease inhibitor with efficacy in a murine infection model, co-authored with Pfizer, Science Advances, Aug 2024. Preclinical. | Public record |
| Pfizer programs | Three agreements: Dec 2020, Jan 2022, Jan 2025. The company says one AI Lab program cleared its first stage gate 40 percent faster than forecast. No compound from the collaboration has been named or entered a trial. | Company and partner |
| Amgen programs | Announced Nov 16, 2023 for up to five small molecule programs. All terms undisclosed. No program named since. | Company-stated |
| Fertility programs sold | EMD Serono acquired the FSHR and LHCGR agonist programs on Sep 9, 2026. The release states both “are currently in preclinical development, and their safety and efficacy in humans have not been established.” Price undisclosed. | Company-stated |
| The $1Bn partnerships line | Used since at least November 2023. Disclosed values reach $610 million with Pfizer; Amgen terms were never disclosed. C&EN reached about $1.3 billion only by counting the NIH center. | Differs from disclosures |
| The largest-grant-in-NIH-history line | Company boilerplate describes the ASAP award that way. The AViDD program totalled about $577 million across nine centers; ASAP’s initial $68 million was split among eight institutions. | Differs from the record |
| FDA records and registered trials | None found under PostEra as of Sep 22, 2026. | Not found |
Read plainly: the platform has a checkable record of producing good chemical matter fast, and two large pharma companies and one buyer have paid for it. What it has not produced is a drug. Every molecule traceable to PostEra sits at preclinical stage or earlier, and the approved medicine DNDi credits to the Moonshot data, Shionogi’s ensitrelvir, was made by someone else who read the open files.
What to watch
- The PMOS candidate. The founders told C&EN in September 2026 they plan to start clinical trials early in 2027, the first PostEra-designed molecule in a human.
- A Series B. No disclosed venture money since January 2022, and the company said in late 2025 it was raising.
- Whether EMD Serono advances the FSHR or LHCGR agonist into the clinic.
- Funding for ASAP-0017445. DNDi is publicly asking donors to pay for Phase I; NIH AViDD money ran to 2025.
- Any named compound out of the Pfizer AI Lab or the Amgen collaboration.
In their words
“We realised there were hundreds of chemists sitting at home, with their projects on hold”
Alpha Lee, CSO, Chemistry World, 2020 · Independent
“Most AI approaches to drug discovery focus on answering the question of what compounds to make, but miss how to make those compounds”
Alpha Lee, Chemistry World, 2020 · Independent
“Think of it almost like taking one piece of jigsaw off and snapping many, many building blocks on”
Alpha Lee, C&EN, November 2025 · Independent
“We are going to try and help 50% of the population. That doesn’t seem niche to us. These are huge conditions, and really, they are chemistry challenges, in our eyes.”
Aaron Morris, CEO, C&EN, November 2025 · Independent
“The status quo in IVF is that women receive a lot of injections”
Alpha Lee, C&EN, September 2026 · Independent
“We believe AI can significantly improve drug discovery but we’re also aware of the hype around the technology.”
Aaron Morris, Amgen collaboration release, November 2023 · Company release
“Here we have a promising agent; it is almost ready to be tested in humans, is royalty-free, and could become a generic treatment.”
Peter Sjö, Head of the Drug Discovery Programme, DNDi, September 2025 · Partner
Related companies
Sources
- Public recordOpen science discovery of potent noncovalent SARS-CoV-2 main protease inhibitors
- Public recordDiscovery of SARS-CoV-2 papain-like protease inhibitors with efficacy in a murine infection model
- Public recordNIAID Announces Antiviral Drug Development Awards
- Public recordNIH awards initial $68M for ASAP
- Public recordASAP-0017445 portfolio page
- Public recordCOVID Moonshot portfolio page
- Public recordForm D, PostEra Inc., CIK 0001904407
- IndependentHow PostEra designs drugs from code
- IndependentMerck KGaA buys PostEra’s oral fertility molecules
- IndependentPfizer doubles down on AI partnership with antibody-drug conjugate deal
- IndependentPostEra points its synthesis algorithm at coronavirus
- IndependentOpen-science approach delivers a promising pre-clinical candidate
- CompanyPostEra’s Fertility programs acquired by EMD Serono, Inc.
- CompanyPostEra announces expansion to $610M with Pfizer
- CompanyPostEra Announces a Research Collaboration with Amgen
- CompanyPostEra announces $260M AI Lab collaboration with Pfizer
- CompanyPostEra website and company page
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
