Aitia
Aitia, the company formerly known as GNS Healthcare, uses causal AI to discover drug targets and biomarkers. Its Gemini Digital Twins are computer models of human disease built by its reverse engineering and forward simulation engine, REFS, from patient multiomic and clinical data, which it runs through counterfactual experiments to find the genes and molecular interactions that drive a disease and markers of which patients will respond. Aitia has built twins in neurodegeneration (Alzheimer's, Parkinson's, Huntington's and ALS), oncology (gliomas, pancreatic cancer, multiple myeloma and prostate cancer), cardiometabolic disease (atherosclerosis, NASH and heart failure with preserved ejection fraction) and lupus nephritis, and it works through research partnerships rather than selling software.
Partners include Servier, for brain cancer and Parkinson's programs; Orion, in oncology; Ono Pharma, for neurological targets from September 2026; Gustave Roussy, on the biology of several cancers; and the Pancreatic Cancer Action Network, which supplies deidentified patient data. Aitia also uses Charles River's Logica discovery service in its own programs. A 2017 study in Lancet Neurology by its scientists identified clinical and genetic predictors of motor progression in newly diagnosed Parkinson's disease and tested them in a second cohort. Colin Hill cofounded the company and leads it from Cambridge, Massachusetts.
Capability Axes
An AI Health Index grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check. How grades read
Aitia's REFS engine reverse engineers networks of genes, molecules and clinical outcomes from patient multiomic data and then simulates them forward, and the resulting Gemini Digital Twins are what it brings to every partnership: targets, biomarkers and patient response markers come out of in silico experiments on those models. It sells no software and runs no wet lab business of its own that stands apart from the models, so the discovery engine and the partnerships both rest on them. Even the name comes from the Greek word for causality, the idea the company is built around.
The models propose and partners develop, with the review step left implied. Gemini Digital Twins run counterfactual experiments and surface candidate targets and biomarkers, and partners such as Servier and Orion carry programs into discovery and development; Aitia also uses Charles River's Logica service in its own programs. How candidates are reviewed by scientists, what evidence is required before a target is pursued and who decides are not described. Nothing suggests the models act without people, but no oversight structure is set out either, and the leadership team of drug developers is the only sign of one.
REFS, short for reverse engineering and forward simulation, infers causal networks of genes, molecules and clinical outcomes from patient multiomic and clinical data, and the Gemini Digital Twins built with it are run through counterfactual simulations, which Aitia says evaluate trillions of parameters on supercomputing infrastructure. The 2017 Parkinson's study and earlier Huntington's network papers describe the approach in published form. The specific datasets behind each twin, how models are updated and versions tracked, and how results are checked are not stated.
REFS and the Gemini Digital Twins were built in house over two decades at GNS Healthcare and Aitia, so no outside model provider sits behind the outputs. Data partners are named in announcements, among them the Pancreatic Cancer Action Network and Gustave Roussy, drug partners such as Servier, Orion and Ono bring program data of their own, and Aitia uses Charles River's Logica service in its programs. The computing provider, any other data sources and the terms that govern partner data once it arrives are not stated, and no subprocessor list appears.
A 2017 study in Lancet Neurology by GNS Healthcare scientists, now Aitia, identified clinical and genetic predictors of motor progression in newly diagnosed Parkinson's disease and tested them in a second cohort, the one peer reviewed result with external validation. Later work appears as posters: drivers of cognitive decline in Huntington's and Alzheimer's, cytoskeleton genes linked to neurofilament change, and SPI1 as a marker of Alzheimer's progression at CTAD 2026.
The homepage says discoveries are validated at the bench and in the clinic, without naming one, and no target from the twins is reported to have reached a clinical trial. Partnerships with Servier, Orion, Ono and Gustave Roussy show demand rather than results.
The twins are built from multiomic and clinical data from leading partners, and the only specific statement about that data is that the Pancreatic Cancer Action Network supplies it deidentified. Which datasets are used, how they are deidentified, whether partners' data is kept separate from Aitia's own models, how long it is retained and who else can reach it are not stated. Guardrails and how a safety event is handled are not described either, though for a discovery platform the more pressing question is the provenance of the patient data the models learn from.
The privacy notice, issued by GNS Healthcare, Inc. doing business as Aitia and last updated December 2023, covers site visitors and other contacts, says trial participants receive a separate notice, and never mentions HIPAA. Aitia builds its models from patient multiomic and clinical data supplied by partners, and its Pancreatic Cancer Action Network partnership describes that data as deidentified, but no statement says whether Aitia ever receives identifiable data, acts as a business associate or relies on deidentified data throughout. No position on patient data is taken anywhere outside that website notice.
No SOC 2, ISO 27001 or HITRUST certification is named, no encryption, access control or penetration testing is described, and the site has no trust center or security page. The privacy notice says only that website information is shared with service providers and others. For a company that holds partners' patient multiomic datasets and pharma collaborators' program data, nothing independent vouches for how it is protected. Partners such as Servier and Ono would set security terms in their own agreements, which are not public, so an outside reader has no way to judge them.
No device claim is made, and the platform sits upstream of regulation. Aitia's models generate targets, biomarkers and hypotheses for drug programs, which reach regulators only as drugs and diagnostics developed by Aitia or its partners under their own filings. The site makes no regulatory claim for the twins and does not suggest they are used in patient care.
A model predicting amyloid PET positivity from blood biomarkers, presented in 2023, would need its own regulatory path if it became a diagnostic, and nothing says one is planned. The privacy notice mentions trials Aitia sponsors, without naming any.
Testing across patient groups goes unmentioned. Twins built from patient cohorts inherit the makeup of those cohorts, and no results are given for how findings hold across ancestry, sex, age or disease subtype, and the training datasets are not described by population. Aitia stresses that causal models reveal mechanism rather than correlation, which aids interpretation but is not an evaluation.
No governance process, review of model changes or outside audit is mentioned, and the Scientific Advisory Board section of the site shows placeholder names. A target that holds in one population may not hold in another.
A wrong model has no stated consequence for Aitia. A false target or biomarker from a digital twin would cost a partner research time and money rather than harm a patient directly, and the remedy would sit in each collaboration agreement, none of which is public. No accuracy figures for the twins, no published error rates for target predictions and no commitment tied to output appear. The site has no terms of service, so even the terms for its website are absent. Partners share risk through joint discovery rather than buying a product with a warranty, which is how the business is set up.
No integration with clinical systems exists or is claimed. Aitia works with research datasets assembled from partners such as the Pancreatic Cancer Action Network and cancer centers, and its outputs go to drug discovery partners rather than into care. No EHR, data standard, research platform connector or marketplace listing is named, and nothing describes how partner data is received or formatted.
For a discovery company this is expected, and it means a buyer has nothing to connect to beyond the partnership itself. Clinical data enters only as part of research cohorts that partners assemble and transfer.
Aitia refers to high performance and supercomputing resources without naming a cloud provider, data center or region, and its privacy notice says website data may be processed in countries other than the visitor's. Where the models run and where partner data rests are not stated.
Partners such as Gustave Roussy in France and Servier and Orion in Europe raise the question of where patient data from their cohorts is processed, and no statement addresses it or describes how one partner's data is kept apart from another's. European patient data would carry its own transfer rules, which are not mentioned.
Aitia works through research collaborations, and its announcements show the shape of them: Servier for glioma and Parkinson's programs, Orion in oncology, Ono Pharma for neurological targets, Gustave Roussy on cancer biology, and the Pancreatic Cancer Action Network as a data partner. Financial terms, such as upfront payments, milestones or royalties, are not disclosed for any of them, and no price, software license or data product is offered. A drug maker can tell it would be entering a joint discovery partnership, and nothing about cost, ownership of resulting targets or exclusivity by disease area.
Twins cover Alzheimer's, Parkinson's, Huntington's and ALS, gliomas, pancreatic cancer, multiple myeloma and prostate cancer, atherosclerosis, NASH and heart failure with preserved ejection fraction, and lupus nephritis. The 2017 Lancet Neurology study supports the Parkinson's work, posters cover Huntington's and Alzheimer's, and the SPI1 progression marker presented at CTAD 2026 extends the Alzheimer's line.
The oncology, cardiometabolic and immunology twins have partners or announcements but no published results on the site, so neurodegeneration is where the record is deepest and most of the other named coverage has yet to show results.
Alternatives to Aitia
Other products that do the same kind of work as Aitia, and what each one does.
Valo Health uses AI on large scale human data to find drug targets and design small molecules, through its Opal Computational Platform.
Molcure designs antibodies and peptides for drug companies, using a large language model for antibodies that it built from scratch together with physics based simulations and experimental data from its own biolab.
Bexorg tests drugs for central nervous system diseases in donated postmortem human brains before they reach clinical trials, working for pharmaceutical and biotechnology sponsors and on its own licensed candidates.
insitro is a South San Francisco AI therapeutics company founded in 2018 by Daphne Koller, built on causal biology.
Clinical stage biotechnology company now trading as Verge Labs, applying machine learning to human tissue derived multiomic data in neurology.
Insilico Medicine is a generative AI drug discovery company (HKEX: 3696), headquartered in Cambridge, Massachusetts with operations in Hong Kong, Abu Dhabi and Shanghai, that both licenses software and develops its own medicines.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | BAA Tier | Implementation | Source |
|---|---|---|---|---|
|
Not published.
|
Aitia works through research collaborations with drug makers, cancer centers and patient foundations rather than selling software or data. Financial terms of its collaborations with Servier, Orion, Ono and others are not disclosed. | Not published. | None stated. | Vendor Published |
No upfront, milestone or royalty terms are disclosed for any partnership.