AI multi-biomarker blood tests: a research framework summary
By OncoFirm Diagnostics · Last updated October 2026
AI multi-biomarker blood tests combine several circulating signals, such as DNA, proteins and immune markers, and use a trained model to turn them into one risk estimate. The approach is scientifically plausible and clinically unproven: a model that separates known cancer cases from healthy controls has not thereby been shown to help patients.
This page summarizes a research framework prepared by OncoFirm™ Diagnostics. It is a narrative summary of published concepts. It reports no new patient data and no OncoFirm performance data. OncoFirm is developing diagnostic technologies and has a commercial interest in the subject.
Screening and triage are different intended uses
| Population screening | Clinical triage | |
|---|---|---|
| Who is tested | People without symptoms of the cancer in question | People with symptoms, an abnormal finding or a referral |
| Likelihood of cancer before testing | Low | Higher, and dependent on the pathway |
| Main risk | False positives, overdiagnosis and unnecessary procedures | Missed or delayed diagnosis |
| Evidence needed | Prospective screening cohort with long follow-up | The referral population under real workflow conditions |
A model validated for one use cannot be assumed to work in the other.
Biological components
Tumors differ in what they release into blood, and inflammation, infection and benign disease alter many of the same analytes. Panels try to offset this by combining partly independent signals.
Published examples include mutations in cell-free DNA combined with protein markers (Cohen et al., 2018), genome-wide DNA fragmentation patterns (Cristiano et al., 2019) and targeted methylation (Klein et al., 2021). Other candidate classes are tumor-associated antigens, autoantibodies, extracellular vesicles and routine blood counts. See our liquid biopsy explainer and our guide to tumor antigen detection.
Model design
In this framework the algorithm is one component of a test system, alongside the assay and the reader. The run is checked for analytical validity, signals are normalized and risk is then estimated, with quality control kept separate from risk classification.
Three properties matter more than model type: calibration, so that a predicted risk matches the observed frequency; an explicit indeterminate category for uncertain or out-of-range inputs; and a locked version evaluated on data it has never seen. General pitfalls are covered in AI in cancer diagnostics.
Analytical validation first, then clinical validation
Analytical validation establishes that the measurement itself is reliable: precision, detection limits, interference, specimen stability, reader-to-reader agreement and, for multiplex strips, cross-reactivity between antibody pairs.
Clinical validation then tests the locked system in the intended-use population against an accepted reference standard, with stage-specific and cancer-specific analyses. Prospective studies that returned results to participants show how much depends on the diagnostic workup after a positive test (Lennon et al., 2020; Schrag et al., 2023). Clinical utility, meaning whether using the test improves outcomes, is a third question with its own trial-design problems (Minasian et al., 2023).
Bias and human oversight
Known risks include spectrum bias, data leakage, under-represented subgroups, automation bias and false reassurance after a negative result. The framework treats the output as decision support reviewed by a clinician. It adopts TRIPOD+AI for model reporting (Collins et al., 2024), CONSORT-AI for trials (Liu et al., 2020) and the Good Machine Learning Practice principles of the FDA, Health Canada and the MHRA (2021).
A staged development framework for an investigational platform
For an investigational OncoFirm platform, comprising a multiplex fluorescent lateral flow assay, a reader and a locked algorithm, the framework proposes starting from a narrow intended use and passing through defined stages. It is a plan, not a record of completed work.
- Target product profile: population, intended use, specimen, setting and comparator.
- Analytical feasibility: measurable signals and a stable assay configuration.
- Biomarker and model discovery in retrospective specimens.
- Analytical validation of assay, reader and software together.
- Independent clinical validation of the locked system.
- Prospective silent-mode evaluation, with results not influencing care.
- Prospective interventional evaluation of safety and clinical utility.
- Regulatory review and controlled deployment.
Claims this framework treats as unsupported until shown
Each statement below is a testable performance or utility claim. The framework holds that none should be made for any test, OncoFirm’s included, without evidence designed to support it.
- That a blood test establishes a diagnosis of cancer.
- That a negative result rules out cancer.
- “Ultra-early detection”.
- Accuracy equivalent to a central laboratory.
- “Higher sensitivity” than existing methods.
- That testing prevents cancer deaths.
- That testing reduces unnecessary procedures.
- That multiple cancers are detected from a single drop of blood.
Acceptable wording before such evidence exists: investigational, under development, intended for prospective validation.
Where research partners fit in this framework
The early stages need partners: clinical sites and biobanks with well-characterized specimens, laboratories able to run reference immunoassays, and biostatisticians to prespecify analyses. Open questions are listed under research and development priorities; to propose joint work, use the collaboration opportunities page.
Frequently asked questions
Is this a clinical study?
No. It is a narrative research framework that reports no new patient data and no OncoFirm performance data.
What is the difference between screening and triage?
Screening tests people without symptoms; triage helps prioritize people already under investigation. They need different thresholds, study designs and endpoints.
Does OncoFirm have a multi-cancer blood test?
No. The platform described is at the research stage and under development, with no established analytical or clinical performance.
Related research and resources
- The future of cancer screening — the open evidence questions.
- Clinical development pathway — how analytical and clinical studies are planned.
- Discuss a grant opportunity — propose a validation study.
References and further reading
- Cohen JD, Li L, Wang Y, et al. Detection and localization of surgically resectable cancers with a multi-analyte blood test. Science. 2018;359(6378):926–930.
- Cristiano S, Leal A, Phallen J, et al. Genome-wide cell-free DNA fragmentation in patients with cancer. Nature. 2019;570(7761):385–389.
- Klein EA, Richards D, Cohn A, et al. Clinical validation of a targeted methylation-based multi-cancer early detection test using an independent validation set. Ann Oncol. 2021;32(9):1167–1177.
- Lennon AM, Buchanan AH, Kinde I, et al. Feasibility of blood testing combined with PET-CT to screen for cancer and guide intervention. Science. 2020;369(6499):eabb9601.
- Schrag D, Beer TM, McDonnell CH, et al. Blood-based tests for multicancer early detection (PATHFINDER): a prospective cohort study. Lancet. 2023;402(10409):1251–1260.
- Minasian LM, Pinsky P, Katki HA, et al. Study design considerations for trials to evaluate multicancer early detection assays for clinical utility. J Natl Cancer Inst. 2023;115(3):250–257.
- Collins GS, Moons KGM, Dhiman P, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378.
- Liu X, Cruz Rivera S, Moher D, Calvert MJ, Denniston AK; SPIRIT-AI and CONSORT-AI Working Group. Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension. Nat Med. 2020;26(9):1364–1374.
- U.S. Food and Drug Administration, Health Canada, MHRA. Good Machine Learning Practice for Medical Device Development: Guiding Principles. 2021.
Research and regulatory status
OncoFirm™ technologies described on this site are in research and development. Their performance characteristics have not been established, and no OncoFirm product has been cleared, approved or authorized by the U.S. Food and Drug Administration. Nothing here is medical advice. Reference to a federal agency or funding program does not imply endorsement, funding, sponsorship or affiliation.
