Consumer genomics · Third party DNA interpretation
Genomelink
A Berkeley and Tokyo company that takes the DNA file you already bought from somebody else and returns hundreds of trait reports. Two peer reviewed papers sit under it. Its own terms of use say the discoveries it reports have not been clinically validated.
The file in your downloads folder
If you have taken a DNA test from 23andMe, AncestryDNA or MyHeritage, there is a text file you can download. It holds a few hundred thousand single letter positions, the ones the genotyping chip happened to read. The company that sold you the kit used only some of them.
Genomelink exists to read the rest. You upload, and the site returns trait reports: caffeine metabolism, endurance, extraversion, memory, vitamin B12, facial wrinkles. The free tier gives 100. The paid library, checked live by KnowYourDNA in May 2026, runs to 352 or more, with an unlimited plan promising up to 500.
The company calls this a DNA App Store. It owns no lab, sells no kit and collects no saliva. It owns the interpretation layer on top of somebody else’s chip.
What a trait report actually is
Nearly every report is a genome wide association study rendered as a dial. A study finds that people carrying a particular letter at a particular position score slightly higher on some measure. Genomelink checks your file and shows where you land.
The company is more candid about this than most. Each trait carries a Scientific Reliability Score, a four point scale adding a point each for a sample of 10,000 or more, statistical significance, independent replication, and meta analysis. A reviewer at Gene Food found his extroversion result rested on four positions and was labeled suggestive.
What the dials cannot do is carry weight per person. Robert Plomin and Sophie von Stumm, two researchers broadly optimistic about polygenic prediction, set the ceiling in Molecular Psychiatry in 2021: the best behavioral scores predict about 7 percent of the variance in general cognitive ability and 11 percent in years of schooling, and even 10 percent is a correlation near 0.32. Real signal in a population, a wide cloud around any one person.
Genomelink sells intelligence, personality and career fit reports against that arithmetic. Its marketing page tells readers that being an entrepreneur might be in their DNA. The National Human Genome Research Institute’s guidance for clinicians is the opposite register: the usefulness of nutrition, fitness and lifestyle recommendations from consumer genetic testing is, in its words, unclear in most cases.
Four founders and a flight home
Tomohiro Takano studied at International Christian University in Tokyo, helped set up Impact Hub Tokyo, and joined the medical platform company M3 in 2015 to work on a new genomics business. Three of the four founders were friends from ICU or friends of those friends. Yuta Matsuda had worked on MyCode, a large Japanese consumer genetic test. Kensuke Numakura came from a Sony and Illumina joint venture. They built on weekends from 2016 and incorporated in the United States in 2017, choosing the American market because, as Takano put it, roughly 40 times more people there had their own DNA data than in Japan.
The first year nearly ended it: no visa, limited funds, a product that did not take off. The Berkeley SkyDeck acceptance email arrived, by his account, one day before they were due to fly home and possibly shut down. SkyDeck led to Y Combinator’s Summer 2021 batch, and to Carlos Bustamante, a Stanford professor and 2010 MacArthur Fellow, joining as board observer and head of science advisory.
The part that became research
Bustamante’s arrival produced the strongest thing on the record. In 2022, Human Genomics published a paper with Genomelink and Awakens Japan affiliations describing exactly what the platform is: users upload genotype files and answer health questionnaires, and those answers become a research cohort. It reported 4,550 participants for type 2 diabetes and 4,528 for hypertension, recovered 164 of 272 previously reported genome significant variants, and built risk scores reaching an area under the curve of 0.68.
A second paper in 2024 used 8,417 people from the Biobank of the Americas-Genomelink to validate polygenic models, and quantified how badly European derived scores travel: performance fell to 51.3 percent of the European benchmark in South Asians, 46.6 percent in East Asians and 39.4 percent in Africans. That is a real contribution, and also a plain statement of what uploaded consumer genomes are worth to whoever holds them.
A clinical claim appears
In 2025 and 2026 the company began pointing at T1D Scout, a collaboration with the advocacy site T1D Strong led by co-founder Yuta Matsuda. It is a saliva genetic risk score meant to flag children at elevated risk of type 1 diabetes before autoantibody testing. The T1D Strong page says the test is available in 2026 and currently undergoing FDA approval process, and reports a pilot in which more than 3,000 families signed up and over 800 samples came back.
No matching record exists in the FDA 510(k), De Novo or premarket approval databases, and no study appears on ClinicalTrials.gov. Those searches returned nothing under Genomelink, Awakens or T1D Scout on September 23, 2026.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| Peer reviewed work | Two papers in Human Genomics, Sep 2022 and Sep 2024, listing Genomelink, Inc., Berkeley and Awakens Japan K.K., Tokyo. Takano and Bustamante author both. AUC 0.68 for the 2022 cardiometabolic scores; 8,417 people used for validation in 2024. | Public record |
| FDA status | No 510(k), De Novo or PMA record found under Genomelink, Awakens or T1D Scout, searched Sep 23, 2026. FDA states that tests for non-medical, general wellness or low risk purposes are generally not reviewed before they are offered, naming athletic ability prediction as an example. | Public record |
| T1D Scout regulatory claim | T1D Strong markets the test as available in 2026 and undergoing FDA approval, and reports a risk score right about 90 times in 100. No FDA submission and no registered trial were found. | Claim without a record |
| Marketing versus terms | The DNA Reports page promises results you can trust and tells readers their personality report holds the key to their success. The Terms of Use, effective Sep 14, 2026, say the discoveries have not been clinically validated and the technology is not appropriate for clinical testing. | Differs from marketing |
| Company accuracy evidence | The published accuracy post is a user vote: 223 of 300 traits, 74.3 percent, rated accurate by a majority of users. The five rated least accurate were acne vulgaris, alcohol drinking behavior, peach allergy, word reading ability and views on environmentalism. User agreement is not clinical validity. | Company-stated |
| Privacy commitments | Policy updated Aug 28, 2026, effective Sep 14, 2026. Deletion within 30 days for anyone, anywhere, no reason required. Genetic data to law enforcement only on a warrant specifically authorizing it. An absolute ban, regardless of consent, on sharing with insurers, employers or risk assessment businesses. | Company document |
| Where genomes can still go | The DNAsolves opt-in sends genetic data to Othram, Inc. for forensic genealogy. The policy says the company may receive compensation from Othram, so the transfer may count as a sale under some state laws. Maryland, Florida and New Jersey residents are excluded. | Company document |
Read plainly: the company has done real science with the data its users hand over, and its consumer documents are more protective than the category norm. It also sells intelligence and personality reports whose predictive power, on the friendliest published account, is a correlation near 0.32, and markets a diabetes test as under FDA review with no visible FDA file.
What to watch
- Whether T1D Scout appears in an FDA database or on ClinicalTrials.gov. A risk score sold to parents of young children is not a general wellness product.
- Whether the bankruptcy clause is ever tested. After 23andMe, the clause requiring any acquirer to adopt the same policy is the most consequential sentence Genomelink has written.
- A first SEC Form D, or an announcement specific enough to reconcile the four totals the commercial databases carry.
In their words
“We started Genomelink in 2016 as a weekend project between myself and my cofounders, Yuta and Ken.”
Tomohiro Takano, co-founder and CEO, Berkeley SkyDeck interview, 2025 · Interview
“Your personality report by Genomelink holds the key to your success.”
Genomelink DNA Reports page, accessed Sep 2026 · Company marketing
“The genetic discoveries that we report have not been clinically validated, and the technology we use, while used by the research community, is not appropriate for use in clinical testing.”
Genomelink Terms of Use, effective Sep 14, 2026 · Differs from marketing
“you may learn that you are predisposed to sadness, anger, or loneliness”
Genomelink Informed Consent, on the risks of using the service · Company document
“Genomelink offers the easiest process for deleting raw data I have seen from any similar site.”
John O’Connor, founder of Gene Food, product review, 2023 · Independent
“Most of those models were designed for research, not for families.”
Yuta Matsuda, co-founder and CEO of T1D Scout, quoted by T1D Strong, 2026 · Independent
Related companies
Sources
- Public recordValidating and automating learning of cardiometabolic polygenic risk scores, PMID 36076307
- Public recordPolygenic risk score portability across diverse populations, PMID 39218908
- Public recordFDA and SEC EDGAR database searches for Genomelink and Awakens, no results
- Public recordDirect-to-Consumer Tests
- Public recordProvider FAQ on direct-to-consumer genetic testing
- Public recordAttorney General James sues 23andMe over genetic data
- IndependentPolygenic scores: prediction versus explanation
- IndependentGenomelink Review (2026)
- InterviewAgainst the Odds: How One Japanese Startup Got Its Start in the U.S.
- CompanyPrivacy Policy
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
