Drug discovery · Kinase inhibitors
Harmonic Discovery
A New York company using machine learning to design kinase drugs that engage several targets on purpose while avoiding the ones that cause harm. Five years in, every program is still preclinical.
A pill that copies a gene therapy
Late in 2025, at an event in Abu Dhabi, a FINTECH.TV interviewer asked two people from Harmonic Discovery what they were building. The answer was not a kinase platform pitch. It was sickle cell disease.
There is already a cure, the speaker said, in the form of gene therapy. The problem is the price and the plumbing: hospital infrastructure and a bill in the millions. So the company set out to do the same job with a tablet.
The transcript does not name which of them was speaking. What it records is a date: a phase one trial in 18 to 24 months, and a fundraising round opening early the following year. Neither has surfaced in any public record. That gap between the plan and the paperwork is the fair shape of this company today.
The problem with magic bullets
Kinases move phosphate groups onto other proteins, switching them on, off, or sideways. Hundreds sit at the control points of cell growth and immune signalling, and when they misfire, cancer and autoimmune disease follow. They are also one of the hardest target families to hit cleanly. Every kinase binds the same small molecule, ATP, so their pockets look alike. A drug aimed at one routinely touches a dozen others. Sometimes that helps. Sometimes it produces cardiac toxicity or bone marrow suppression, and the drug that worked in a dish fails a patient.
The conventional response is to chase selectivity: one drug, one target. Harmonic Discovery argues that is the wrong request to make of chemistry that is promiscuous by nature. Its bet is that you can specify the whole profile up front, naming the kinases you want engaged and the anti-targets you want left alone, then let generative models search only that narrow band of chemical space.
Three founders, three trades
Rayees Rahman, the chief executive, worked out the computational core during a PhD in biophysics and systems pharmacology at the Icahn School of Medicine at Mount Sinai, building machine learning methods to describe the three dimensional structures of kinases and their inhibitors.
Marcel Patek, the chief scientific officer, is the counterweight: more than 25 years in drug discovery, director of chemistry at Sanofi, vice president of chemistry at Icagen. Jason Lee, the third co-founder, came through Deutsche Bank and OrbiMed before earning an MD. The advisory board is a roll call of kinase and leukaemia research: Olivier Elemento of Weill Cornell, Omar Abdel-Wahab of Memorial Sloan Kettering, Tero Aittokallio of Oslo and Andreas Bender of Cambridge.
Money, then a molecule
Harmonic launched publicly on September 8, 2022 with $8 million in seed funding led by Innovation Endeavors, alongside Fifty Years, Y Combinator, Boom Capital, Caffeinated Capital and angels including Ethan Perlstein and Harry Glorikian. Joel Dudley of Innovation Endeavors, formerly chief scientific officer at Tempus, joined the board. Fifty Years described that money as following a pre-seed it had also joined.
Seventeen months on, the company did something platform startups often postpone: it bought an asset. In January 2024 it took an exclusive licence to HD-10019, a type II FLT3 inhibitor developed in the laboratories of Hong-Yu Li, then at the University of Arkansas for Medical Sciences, and Neil Shah at UC San Francisco, designed to survive the resistance mutations that defeat approved FLT3 drugs and to spare the bone marrow. Harmonic said its own selectivity models corroborated the profile. In April 2024 it added a discovery collaboration with Turbine, a company that simulates cell signalling, aimed at NEK1, a so called dark kinase. In October 2024 BioWorld reported that it had presented an FLT3 inhibitor with an improved safety profile.
The pivot in the paperwork
Then the stated focus moved. Company releases through 2024 put the lead programs in malignant haematology. A BIO-Europe profile that year widened it to malignant and benign haematology. By September 2, 2025, when Hub71 in Abu Dhabi announced Harmonic Discovery in its seventeenth cohort, the accelerator described a lead programme targeting sickle cell disease and put total funding at $8.5 million. Sickle cell is benign haematology, not oncology. Whether HD-10019 is still the lead asset, a parallel one, or shelved, no public document says.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| Platform paper | Nature Communications, Aug 31, 2024. Models trained on both single dose and dose response assay data, then wet lab profiling of 347 compound kinase pairs, “achieving a high hit rate of 40% and a negative predictive value of 78%”. The paper cites conventional virtual screens at 5% to 25%. | Public record |
| Other publications | Bioinformatics Advances 2023 on AlphaFold2 kinase conformations; a Current Opinion in Structural Biology 2024 review of AI for targeted polypharmacology; a preference optimization paper for molecular language models, Oct 2023; a medRxiv preprint of Aug 2024 linking predicted kinase binding to adverse event reports. | Public record |
| HD-10019 | Licensed exclusively from BioVentures at UAMS and from UCSF, announced Jan 30, 2024, with in vivo results described as indicating a potentially best-in-class profile. Terms not disclosed. | Company-stated |
| IND-enabling work | The January 2024 release said IND-enabling studies were expected to commence in late 2024. No public confirmation that they started. | Not found |
| Clinical trials | No study sponsored by Harmonic Discovery found in trial registry records as of Sep 22, 2026. | Not found |
| Federal funding | No award found in NIH RePORTER, by organisation or by principal investigator, and none in NSF records. No SBIR or STTR grant located. | Not found |
| Turbine collaboration | Announced Apr 1, 2024 on the dark kinase NEK1. Harmonic takes the chemistry, Turbine the simulation and wet lab biology. Terms not disclosed, no result reported since. | Partner-stated |
| Lead indication | Company releases in 2024 place the lead programs in malignant haematology. Hub71 in Sep 2025 describes a lead programme targeting sickle cell disease. | Claims differ |
| Total raised | $8 million per company releases from 2022 through 2024. $8.5 million per Hub71, Sep 2025. No Form D on file to settle it. | Claims differ |
Read plainly: the computational claim has been tested in public and held up. A 40 percent experimental hit rate, with negative predictions checked as well as positive ones, is a real result and an unusually candid one for this field. What has not been tested is the thing that matters to patients. No Harmonic molecule has entered a human being, and across the industry roughly 95 percent of programs that reach phase one never reach approval. Harmonic is at the start of that road, not partway along it.
What to watch
- Whether a phase one trial appears on a public registry. The 2025 interview put it 18 to 24 months out.
- Which indication goes first. FLT3 mutated AML and sickle cell disease need different trials, different regulators and different capital.
- A Form D, or a named Series A. The company said a round would open early in 2026, and nothing has been filed.
- Whether the Abu Dhabi base becomes a development site or stays a soft landing.
- Peer reviewed data on HD-10019 itself, which so far appears only in releases and one conference report.
In their words
“There is an enormous opportunity to create safer and more effective drugs.”
Rayees Rahman, CEO and co-founder, launch release, September 2022 · Company release
“Their machine learning models and computational approaches are extremely precise and have been validated experimentally,”
Joel Dudley, partner, Innovation Endeavors, launch release, 2022 · Investor, in company release
“Since we provided the first evidence that FLT3-ITD is a valid therapeutic target in AML in 2012, three FLT3 tyrosine kinase inhibitors have been approved in the USA and throughout much of the world.”
Neil Shah, MD, PhD, UCSF, licensing release, January 2024 · Company release
“We believe that the dark kinase NEK1 holds great potential as a drug target for developing cancer therapies addressing important resistance challenges in the clinic,”
Dániel Veres, MD, PhD, CSO and co-founder, Turbine, April 2024 · Partner release
“We found that fancy AI models are often overfitted and perform poorly, while simpler models do way better.”
Anna Cichońska and Rayees Rahman, company blog on their Nature Communications paper, August 2024 · Company
“so what we are doing at Harmonic is essentially mimicking the mechanism of action of gene therapy but developing a pill that can be easily taken at home orally, so very safe but accessible to everyone.”
Unnamed Harmonic Discovery speaker, FINTECH.TV interview, Abu Dhabi, 2025 · Interview
Related companies
Sources
- Public recordLeveraging multiple data types for compound-kinase bioactivity prediction
- Public recordConformational landscape of AlphaFold2-predicted kinase structures
- Public recordAI for targeted polypharmacology
- Public recordDelineating adverse events of kinase inhibitors
- Public recordNIH RePORTER searches returning no awards
- Public recordEDGAR company search returning no matching companies
- IndependentHub71 welcomes record AI startups in latest cohort
- IndependentHarmonic Discovery presents FLT3 kinase inhibitor with enhanced safety profile
- IndependentFirm advises Harmonic Discovery on launch, $8 million seed round
- InvestorHarmonic Discovery
- InvestorTuning precision polypharmacology: our investment in Harmonic Discovery
- InterviewSickle cell treatment: the journey of Harmonic Discovery
- IndependentThe AI drug revolution needs a revolution
- CompanyHarmonic Discovery launches with $8 million in seed
- CompanyLicensing agreement with BioVentures and UCSF for FLT3 mutated AML
- PartnerTurbine and Harmonic Discovery collaboration on NEK1
- CompanyCompany blog: Introducing Harmonic Discovery, and on AI in drug discovery
- CompanyCompany website, science and team pages
- CompanyBIO-Europe startup spotlight profile
- CompanyY Combinator company page, Summer 2021 batch
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
