Genomate Health

The computational model is the product. The platform calculates a cumulative score for each match between a drug and a genomic profile, ranks therapies targeting cancer driver alterations, and returns a ranked list of drugs most likely to be effective for that individual. Without the scoring model there is no output, only a variant list the oncologist already has.

Among the better designed oversight postures in this index, on two counts. Structurally, the output feeds molecular tumor board recommendations rather than reaching the prescribing decision directly, and clinics in Central and Eastern Europe incorporated it into that workflow. Architecturally, the company describes its tool as open box AI in contrast to black box models, meaning the reasoning behind a ranking is inspectable. For a system telling an oncologist which drug is most likely to work, inspectability is the difference between decision support and an oracle.

The method is described concretely rather than in marketing terms: a knowledge model performing automated data compilation, evaluation, and synthesis, scoring each drug to genomic profile match and producing a cumulative rank. The company's own open box framing is a transparency claim it invites scrutiny against. Peer reviewed and conference publications describe the model's performance. Falls short of an A because no technical specification, feature set, or model documentation is published for independent audit.

Nothing published addresses patient data handling: no retention period, deletion procedure, end of contract disposition, sub processor list, residency statement or position on secondary use and model development was located. Two features of this product make the gap worth pressing rather than filing as ordinary early stage thinness.

The data is genomic, so it sits in the category where de identification is weakest, because a genome retains identifying structure that removing names and dates does not touch. Any commitment about de identified use would mean less here than elsewhere, and the absence of any commitment at all means correspondingly more. The second follows from the company's own central claim by simple inference.

It markets a clinically validated model and has published its method in a peer reviewed journal, which means patient data was used to develop and validate it, and nothing published states where those cohorts came from, on what consent basis, or whether data submitted by a customer joins the corpus used for further development.

A company whose principal claim is clinical validation should be able to describe the provenance of the data behind that validation, since the two are the same asset viewed from different ends. Ask for a retention schedule in writing, ask directly whether submitted patient data is used for model development or research, and establish which jurisdiction processes and stores it given the company operates across two.

Evidence is validated against trial data rather than asserted, which is rare in precision oncology decision support. The company reports that in the SHIVA01 clinical trial, patients who received therapy recommended by its model had a 4X higher positive response rate than those whose therapy was selected on the presence of a single biomarker but not recommended by the model.

A published performance analysis in that trial exists, alongside a real world performance analysis in pediatric tumors and an ASCO abstract evaluating the model's clinical performance in immune checkpoint therapy response prediction for lung cancer. Selection into the Mayo Clinic Platform_Accelerate program adds an independent validation signal. Caveat for buyers: the SHIVA01 analysis is retrospective application to trial data rather than a prospective trial of the tool itself.

Nothing published addresses patient data handling. No retention period, deletion procedure, end of contract disposition, subprocessor list, residency statement or position on secondary use and model development was located.

Two features of this particular product make the gap worth pressing rather than filing as ordinary early stage thinness.

The data is genomic. Tumour sequencing results, combined with the clinical profile needed to rank therapies, sit in the category where de identification is weakest, because a genome retains identifying structure that removing names and dates does not touch. Any commitment about de identified use means less here than it would elsewhere, and the absence of any commitment at all means correspondingly more.

The company also markets a clinically validated model and has published its method in a peer reviewed journal, which means patient data was used to develop and validate it. Nothing published states where those cohorts came from, on what consent basis, or whether data submitted by a customer joins the corpus used for further development. A company whose central claim is clinical validation should be able to describe the provenance of the data behind it.

Buyers should ask for a retention schedule in writing, ask directly whether submitted patient data is used for model development or research, and ask which jurisdiction processes and stores it given the company operates from both the United States and Hungary.

No HIPAA statement, business associate agreement, template or terms were located across two separate searches, and the company does not identify itself as a business associate in any retrieved material.

Some of this is explained by stage and history. This is a small company, with total funding of roughly ten million dollars, which formally entered the United States market in 2025 having operated from Hungary under European data protection law. A vendor whose compliance documentation was built for one regulatory regime will not have a HIPAA position ready on arrival, and that is unsurprising rather than evasive.

It does not remain unsurprising for long. The company is now incorporated in Massachusetts, is marketing to United States oncologists, and receives tumour genomic profiles and clinical data from treating physicians in order to produce its therapy rankings. That is processing patient data on behalf of a provider, which is a business associate relationship, and a United States cancer centre will require the agreement before any clinical use.

Buyers should ask whether the company will execute a business associate agreement, which legal entity signs it given the recent acquisition of its former parent, and how data flows between the United States and Hungarian operations are governed.

No trust centre, security page, certification or third party attestation was located across two separate searches, and no security controls are described in any retrieved material.

The stage explains part of this. A company of roughly ten million dollars in total funding, two years past its own founding and one year into United States operations, will not typically have completed an independent security assessment, and the absence should not be read as a statement about the underlying engineering.

It is nonetheless the axis with the shortest runway here, for a specific reason. The data this product consumes is tumour genomic profiling, which is the most sensitive and least de identifiable category in healthcare, and the customers being pursued are cancer centres and academic institutions whose vendor security reviews are among the most demanding in the sector. An attestation is normally a precondition for moving from pilot to production in that setting, not a later refinement.

Buyers should ask what the company holds today, what assessment is in progress and on what timeline, and whether any certification held by the recently acquired Hungarian entity covers the product and infrastructure being sold in the United States. A certification obtained by one entity in one jurisdiction does not automatically extend to another.

No clearance, authorisation or published regulatory position was located, and the grade records the absence of a stated position rather than a judgement that clearance is required.

The corporate history makes this more complicated than usual and a buyer needs to understand it. Genomate Health was founded in 2022 as a spin out of a Hungarian medical technology company that had been operating since 2003, and in January 2026 the spin out acquired its former parent. That parent describes itself as having developed decision support and diagnostic medical devices and refers to registration of an early precision oncology tool, which points to a European regulatory history. Whether any of those registrations transferred with the acquisition, and whether they cover the product marketed today, is not stated anywhere published. A device registration is held by a named legal manufacturer, so a change of entity is exactly the event that requires it to be reviewed.

On the substance, this is among the clearest decision support questions in the category. The product ranks targeted therapies for an individual patient on the basis of the complete genetic profile of their tumour. Ranking treatment options for a named patient sits close to a treatment recommendation, and the clinical decision support exclusion under the 21st Century Cures Act turns on whether the clinician can independently review the basis rather than rely on the output. The company describes a computational reasoning model and has published its method, which helps, but nothing published states its regulatory position.

Buyers should ask for the current regulatory status in both the United States and Europe, and should ask specifically which entity holds any European registration following the acquisition.

No formal governance program is published, but the open box design is itself a partial answer to the governance question, since an inspectable model can be audited by the clinicians using it. Participation in the INSTAND-NGS4P initiative with clinical testing across three European sites indicates engagement with external standardization. No bias evaluation across populations or tumor types was located, which matters given the model is trained on published research evidence that carries its own representational skew.

The method is described concretely rather than in marketing terms: a knowledge model performing automated compilation, evaluation and synthesis, scoring each drug against a genomic profile and producing a cumulative rank. A described scoring and ranking mechanism is contestable in a way a bare recommendation is not, because an oncologist who disagrees with a ranking can ask which evidence produced a given score rather than being handed an ordered list with no visible basis.

The company's own open box framing is a transparency claim it invites scrutiny against, and peer reviewed and conference publications describe the model's performance, so the claim has been put in front of reviewers. Held at C because the specification is not available for independent audit: no technical specification, feature set or model documentation is published, so a reader can follow the shape of the method without being able to reproduce or challenge any particular score.

No warranty, indemnity or remediation commitment attaches. The output type also raises a question the publications may not reach, since a therapy ranking in oncology competes with a tumour board's judgement and a clinician needs to know what the ranking does with evidence that is contested, sparse or drawn from a different tumour type. Ask how evidence quality is weighted within a score, what the system does when the literature conflicts, and concordance with tumour board recommendations on real cases.

Integration is with the genomics pipeline rather than the chart. The company works with next generation sequencing reporting partners on a joint solution for precision oncology reporting, tested across three European clinical sites, which is meaningful interoperability in the diagnostic workflow. No EHR integration, HL7, or FHIR support was located.

No deployment model, hosting arrangement, cloud provider, infrastructure description, implementation timeline or availability commitment was located, and no data residency commitment of any kind was found.

Residency is the more consequential half here because of the company's structure. It operates from Cambridge, Massachusetts and Budapest, having entered the United States in 2025 and then acquired its Hungarian former parent in January 2026. A transatlantic operation processing tumour genomic data raises a specific and answerable question, namely whether a United States customer's patient data is processed or stored in the European Union, or the reverse, and under what transfer mechanism. Nothing published addresses it in either direction.

Deployment evidence is limited but not absent. The company is a graduate of a major clinical institution's accelerator programme and has participated in two others, which indicates the product has been through structured evaluation, though an accelerator is not a production deployment and should not be read as one. No named customer institution running the product clinically was located.

Buyers should establish where their data is processed, whether a United States only processing option exists, what integration is required to submit sequencing results, and which named institutions are using the product in routine care rather than in evaluation.

No pricing information of any kind was located. There is no list price, no pricing unit, no tier structure, and no indication of whether the product is sold per patient analysed, per institution, per oncologist or as an annual subscription. For a product that produces a discrete output per patient, the unit is the more important of the two unknowns, because a per case charge and an institutional licence produce very different behaviour in a tumour board deciding whether a given patient warrants the analysis.

What is published is corporate rather than commercial. Total funding of roughly ten million dollars, a two million dollar United States investment in 2025, the leadership structure, and the January 2026 acquisition of the Hungarian company from which the business was originally spun out are all disclosed openly. That transparency about structure is worth noting, since a buyer can at least see who they would be contracting with and how recently that has changed.

Buyers should establish the pricing unit before the figure, ask whether the analysis is billable or reimbursable in their market, and ask what happens commercially if a submitted case yields no ranked therapy option, since a product paid for per case should be clear about what constitutes a delivered result.

Single specialty, which is oncology, and within it the scope is targeted therapy and immune checkpoint therapy selection based on molecular profile. Published evidence spans solid tumors, pediatric tumors, and lung cancer immune checkpoint response. Geographic footprint covers Central and Eastern European oncology clinics with a stated United States market entry, so US deployment references are thinner than European ones.