Diagnostics & genomics · Polygenic risk scores
Haplotype Labs
A one person software company founded by the architect who built 23andMe’s platform, selling labs and prevention clinics a hosted way to run polygenic risk models. One $1.1 million Form D, no FDA record, no published validation.
The report that goes nowhere
The pitch starts with a complaint about a file. Genetic risk reports get delivered by the million, and then they sit. The show notes for a March 2026 podcast episode put the premise in a sentence: if you have ever wondered why your DNA report sits in a PDF collecting digital dust instead of guiding your health decisions, this conversation is for you.
Mike Polcari has reason to take that personally. He spent more than fifteen years at 23andMe, from shortly after its Series A through its public listing, as vice president and chief architect. By his own account in the company’s Y Combinator launch post, the systems his teams built delivered more genetic risk reports to more patients than any other in the world. He also says the part that mattered never landed.
Haplotype Labs, incorporated in Delaware in 2024 and run from an address in Orinda, California, is his answer. It sequences nobody and runs no laboratory. It sells software: a hosted platform, the HaploHub, that lets a genetic testing lab, a concierge clinic, a clinical trial or a consumer brand compute polygenic risk scores and run the standard population genetics steps, phasing, imputation and identity by descent, without hiring a bioinformatics team.
What a polygenic score can and cannot do
A polygenic score adds up the small effects of many common variants into one number estimating predisposition to a trait. The open PGS Catalog, run at EMBL-EBI and the University of Cambridge, exists because researchers publish new ones constantly. Haplotype Labs says its platform can apply more than four thousand publicly available models. It does not say which catalog it counts.
The field has momentum. In March 2026, Allelica announced that polygenic risk had been recognized as a risk enhancing factor in the ACC and AHA guidelines for cardiovascular risk assessment, with its multi ancestry coronary artery disease study cited there.
The limits are just as documented. A 2019 Nature Genetics analysis by Alicia Martin and colleagues found that scores available then were several times more accurate in people of European ancestry than in others, a direct consequence of who has been enrolled in genome wide association studies. A 2025 Nature Reviews Genetics review of clinical use lists suboptimal precision, poor transferability across diverse populations, and low familiarity among patients and clinicians as the barriers to broad adoption.
Regulation sits with the customer, not with Haplotype Labs. A CLIA laboratory reporting disease risk does so as a laboratory developed test, a category the FDA tried to bring under device rules and then lost: on March 31, 2025 a federal judge in Texas vacated the agency’s laboratory developed tests final rule in its entirety. A consumer brand can ship the same number as a general wellness report, outside FDA review entirely. That is the route 23andMe took in 2019 for its type 2 diabetes polygenic report, which drew immediate criticism from epidemiologists about accuracy across ancestries.
The architect
Polcari holds a BS in computer science from Cornell and an MS in biomedical informatics from Stanford. He worked at Merrill Lynch, was an early engineer at Salesforce, then joined 23andMe. The patent record backs the technical claim: he is a named inventor on 23andMe filings covering genetic data storage formats, cohort assembly with privacy protection, genome sharing and collaborative family medical history. One application, number 20250266129, published August 21, 2025 and titled Machine Learning Platform for Polygenic Models, names him first and traces through continuations to a provisional filed May 27, 2020, while he was still at 23andMe. The company he left has since been through bankruptcy: on June 30, 2025 a court approved the sale of substantially all of 23andMe’s assets to TTAM Research Institute, the nonprofit led by co-founder Anne Wojcicki, for $305 million.
Two years, one employee
The timeline is short. Winter 2024 batch at Y Combinator. First securities sale March 29, 2024. Form D filed April 16, 2024 reporting $1,100,000 offered and sold to five investors, with Polcari the only related person. A podcast interview in February 2025. A second in March 2026, framed around building the Hugging Face of multi omics. A slot in the AI and data sciences showcase at PMWC 2026 Silicon Valley, held March 4 to 6, 2026, where the talk title had shifted from a SaaS platform for labs to HaploHub: The AI First EHR for Multi-Omic Precision Health. Through all of it, every listing that reports headcount still reports one person.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| SEC Form D | Filed Apr 16, 2024, CIK 0002017857. First sale Mar 29, 2024. $1,100,000 offered, $1,100,000 sold, $0 remaining, five investors, minimum $50,000. Michael Polcari the only related person, as executive officer, with no directors listed. | Public record |
| Patents | Polcari named inventor on multiple 23andMe filings. Application 20250266129 on polygenic scoring claims priority to a May 27, 2020 provisional. Granted patents in the family list 23andMe entities as assignee. No patent assigned to Haplotype Labs found. | Public record |
| FDA records | No 510(k), de novo or PMA lists Haplotype Labs as of Sep 23, 2026. | Not found |
| Laboratory status | Describes itself as software for laboratories, not as a laboratory. No CLIA certificate in its name found, and none claimed. | Not found |
| Trials, papers, grants | No registered trial, no PubMed record naming the company or platform, no NIH or NSF award. | Not found |
| Named customers | None announced. Buyers described only in categories: testing labs, concierge practices, trials, consumer brands, payers and providers. | Not found |
| Model and cost claims | More than four thousand publicly available models, on any genome wide germline or methylation assay, and 50% to 90% lower sequencing cost through low pass sequencing with imputation. No validation data published. | Company-stated |
| Ancestry portability of PRS | Peer reviewed work from 2019 through 2025 finds scores perform worse outside European ancestry populations, and reviews name transferability as a barrier to adoption. | Public record |
| Headcount | One employee, reported by Y Combinator in 2024 and still reported as one by Latka in 2026. | Independent |
Read plainly: the founder’s track record is documented in the patent file and the technical thesis is coherent, but the company itself has almost no external record. One financing, one person, no customer, no benchmark, no paper, no regulatory footprint. Everything about HaploHub that could be checked by someone other than the company remains unchecked.
What to watch
- A second Form D. None has been filed in the 29 months since the first sale.
- A first named customer, in any of the categories the company lists.
- Any published benchmark comparing model performance across the platform.
- Whether the positioning settles. The 2024 pitch was a scoring platform for labs; the 2026 pitch is a multi omics electronic health record.
- Whether headcount ever moves past one.
In their words
“The technology that my team and I developed at 23andMe has delivered more genetic risk reports to more patients than any other in the world, but the translation to patient outcomes remains unsolved.”
Mike Polcari, founder and CEO, Y Combinator launch post, 2024 · Company
“We’re building the tech I wish I’d had at 23andMe and making it broadly available to the clinic.”
Mike Polcari, Y Combinator launch post, 2024 · Company
“However, it is difficult to correctly and reliably compare the performance of these models or deploy them at scale.”
Mike Polcari, on published polygenic models, Y Combinator launch post, 2024 · Company
“We are building a shared EHR to store and interpret the ’omics cloud and invite each of you to contribute to it.”
Mike Polcari, talk abstract, PMWC 2026 Silicon Valley · Company
“Ask ChatGPT for a PRS score and it’ll confidently give you a wrong answer.”
The DNA of Things, episode 100 notes summarizing Polcari, March 2026 · Interview
“You’re rolling it out to millions of people, but there is a lot we don’t know.”
Peter Kraft, epidemiologist, Harvard, on consumer polygenic scores, MIT Technology Review, 2019 · Independent
“These risk scores do not perform in blacks, disease after disease.”
James Meigs, endocrinologist, Massachusetts General Hospital, MIT Technology Review, 2019 · Independent
“those available today are several times more accurate in individuals of European ancestry than other ancestries”
Alicia Martin and colleagues, Nature Genetics, 2019 · Peer reviewed
Related companies
Sources
- Public recordForm D, Haplotype Labs, Inc., CIK 0002017857
- Public recordUS application 20250266129, Machine Learning Platform for Polygenic Models
- Public recordClinical use of current polygenic risk scores may exacerbate health disparities
- Public recordClinical use of polygenic risk scores: current status, barriers and future directions
- Public recordFederal district court vacates FDA’s laboratory developed tests final rule
- Public recordPGS Catalog, open database of published polygenic scores
- Independent23andMe thinks polygenic risk scores are ready for the masses, but experts aren’t so sure
- IndependentHaplotype Labs profile, funding and investors
- InterviewEpisode 46: The Evolution of Consumer Genomics and What’s Next with Mike Polcari
- InterviewEpisode 100: From 23andMe to the Future of Personalized Genomics
- IndependentPMWC 2026 Silicon Valley event listing
- CompanyHaplotype Labs launch post, profile and job listing
- CompanyHaplotype Labs home, about and contact pages
- CompanyMike Polcari speaker profile and talk abstract
- Company23andMe receives court approval for sale to TTAM Research Institute
- PartnerPolygenic risk score recognized in the 2026 ACC/AHA guidelines
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
