Panakeia Technologies

The models are the product. Panakeia's products infer molecular status from routine H&E images: PANProfiler Colorectal reports microsatellite instability and mismatch repair status, and PANProfiler Breast reports ER, PR and HER2 status, results that otherwise need a separate laboratory assay; the research platform extends the same inference to more than 30 cancer types for drug development. The company sells no viewer, scanner or image management system, and a pathologist cannot read these results off the slide by eye. Remove the models and nothing remains.

The artificial intelligence is the product.

The software declines cases it cannot call, and the path after a wrong call is unstated. In the published colorectal validation PANProfiler returned a result for 86.55 percent of cases, leaving the rest to laboratory testing, and the company describes the product as enabling earlier triage for confirmatory testing where required. It also presents the breast product as moving beyond a screening aid to standalone molecular profiling with a high test-replacement rate, and calls it repeatable and reproducible, which describes consistency rather than correctness. A pathologist cannot check a molecular call by looking at the slide, and nothing states which results should still be confirmed by a wet-lab test or how a false negative would come to light.

The oversight structure described, with the handling of a wrong call missing.

The approach is published in peer reviewed work, with no product version or update practice. The company describes an end-to-end, explainable learning approach trained on large population data and publishes its method in peer reviewed papers, including Communications Medicine (2024) for the platform across cancer types and npj Digital Medicine for the colorectal product. No product version, training data composition or retraining and change notification practice is published for the marked devices.

Approach named without the version and update discipline.

An in-house build, with the operating chain unstated. Panakeia develops its own models on its proprietary platform, so no outside model provider sits between a slide and a result, and it names its academic partners, including the University of Leeds, King's College London, Cancer Research UK and the PathLAKE consortium. Nothing states where the software runs when deployed in a hospital, whether images are sent to Panakeia, which subprocessors apply, or where the training data came from and under what agreements.

A structurally short chain at the model layer, with the operating chain unnamed.

Independent, blinded, multi-site validation is published in a peer reviewed journal. In npj Digital Medicine, a study led by the University of Leeds and Leeds Teaching Hospitals NHS Trust tested PANProfiler Colorectal on 3,576 images from 1,243 colorectal cancer patients at three UK institutions (Leeds, Belfast and St Andrews), with differing tissue preparation and scanners: it returned a result for 86.55 percent of cases, agreed with laboratory MSI and MMR testing in 93.83 percent, and identified 92.54 percent of MSI-high or dMMR cases and 94.02 percent of stable cases. A separate multi-site study supports the breast product, and the platform is described in Communications Medicine (2024). The validation is retrospective on archived cases, and results are reported against the tests the product would replace rather than patient outcomes.

Independently evaluated, peer reviewed, multi site performance with the method available.

Nothing published describes how patient images are handled. The privacy policy covers the website only and says the company collects no health data. The models are described as trained on petabytes of population-scale data, which for this task means slides paired with molecular results for the same patients, but no source institutions, consent or research basis, or de-identification standard is published, and nothing says whether hospital images analyzed by the deployed products are retained or used for development. No safety engineering is published.

Nothing published on how protected information moves through the products.

No position is published and no privacy document reaches the product. The only privacy policy covers the website and states that the company collects no health data. The clinical products are deployed in UK hospitals and CE marked for the EU, so the relevant regime is UK and EU data protection, and no processor position or data processing agreement for patient images is published; the products are not available in the United States.

No statement of status and no privacy document that reaches the product.

Nothing independent is published. The site publishes a privacy policy covering the website and no security or trust page. The company refers to a diagnostic level quality system without naming a certification, and there is no ISO/IEC 27001, Cyber Essentials, NHS Data Security and Protection Toolkit status, SOC 2 report or penetration testing statement.

UK and EU marks for the clinical products, with the US position stated. PANProfiler Breast is UKCA and CE marked as an in vitro diagnostic and, the company states, not yet available in the United States; the company describes two UKCA marked clinical products, colorectal and breast, deployed in hospitals, and labels everything else research use only. It takes part in the MHRA's AI Airlock program and the Friends of Cancer Research Digital PATH project with FDA. FDA's databases list no Panakeia device.

Non-US regimes named, with the marketing matching them.

Independent testing across sites and scanners, with no comparison by patient group. The colorectal validation in npj Digital Medicine was run blind by the University of Leeds across three UK institutions with different populations, tissue preparation and scanners, and the breast validation reports consistent performance across sites and scanner types; the company also takes part in the Friends of Cancer Research Digital PATH project with FDA, which compares multiple vendors' HER2 models against pathologists. No results are presented on the company's pages by patient age, sex, ancestry or tumor subtype, and inferring molecular status from morphology could vary with any of them.

Results within the one group where performance is most at risk, variation across sites and scanners.

A published error rate the company stands behind, and nothing contractual. The colorectal validation in npj Digital Medicine, led by the University of Leeds and announced by Panakeia with its figures, reports on 1,243 patients that the software identified 92.54 percent of MSI-high or dMMR cases and 94.02 percent of stable cases, and declined to call 13.45 percent. A missed MSI-high result can mean a patient is not offered immunotherapy and Lynch syndrome goes unrecognized, with consequences for relatives, and the reviewing pathologist cannot catch it by looking at the slide. The products are UKCA marked, with the manufacturer's post-market duties that carries. It publishes no warranty, indemnity or remediation commitment.

The error rate comes with its method and denominator.

Integration is claimed with no system named. The company says its products integrate with existing digital workflows and require no workflow change, but names no image management system, laboratory information system, report format or standard through which results reach the pathology report.

Integration claimed with no system named.

Nothing published about where the software runs. The company says the products integrate with existing digital workflows and return results in minutes, but states no hosting model, region, on premises option or retention period, for the hospital products or the research platform used by pharmaceutical customers.

Nothing published about where the system runs or where data rest.

No price is published, and how each line is sold can be established. The clinical products are sold to hospitals in Great Britain and the EU through a contact route for medical devices, and the research platform is sold to biopharma as a customizable service with the company's AI, data science, pathology and regulatory expertise and access to its clinical partner network. The company says its testing is faster and cheaper than laboratory methods, with no figure. No per test, per case or platform price is published.

The sales posture is discoverable, with no price.

Two clinical indications with validation behind them, and a research platform without. PANProfiler Colorectal (MSI and MMR) and PANProfiler Breast (ER, PR and HER2) are validated in multi-site studies and deployed in UK hospitals; the research platform is described as profiling thousands of biomarkers across more than 30 cancer types for drug development, with proof of concept published and no clinical validation. The products are not available in the United States.

Coverage named with validation behind part of it.