What is a companion diagnostic?

A specific, regulator-approved biomarker test required for the safe or effective use of a paired therapeutic product, distinct from a general research-use biomarker assay. Developing one from a spatial biomarker requires locking down every assay parameter, then demonstrating both analytical validity, that the assay measures consistently, and clinical validity, that the measurement predicts the outcome it claims to.

How do you develop a spatial companion diagnostic?

By moving a research-stage spatial signature through a defined sequence: locking down every staining, imaging and analysis parameter so the assay stops changing, analytically validating that the locked assay performs consistently across sites and conditions, and clinically validating that its readout predicts patient outcome, typically alongside the paired drug’s own clinical trial.

What validation does a CDx need?

Both analytical validation, confirming the assay measures its target accurately and reproducibly, including across independent laboratories, and clinical validation, confirming that measurement predicts the relevant patient outcome. For spatial biomarkers specifically, multi-institutional studies have demonstrated that even spatial-specific measures such as cell-to-cell proximity, not just abundance, can be standardized across sites through coordinated harmonization.

Spatial biomarker to companion diagnostic

Highlights in this article

What a companion diagnostic requires
Locking down the assay
Analytical validation
Clinical validation and co-development
Standardization and reproducibility hurdles

Walking the companion diagnostic development path from a discovered spatial signature to an assay a pathologist can run reliably in an independent laboratory means confronting a specific engineering problem most biomarker discovery work never has to solve: a finding that held up beautifully in the lab that discovered it has to keep holding up when a different technician, on a different instrument, in a different building, runs the same protocol months or years later.


Key takeaways A discovery-stage assay and a clinical-trial-ready assay are not the same thing; moving between them, "locking down" the assay, is its own distinct development stage. A six-site study achieved robust concordance not only for cell density but specifically for coexpression and proximity metrics, the harder, spatial-specific version of the reproducibility question. A separate multi-institutional effort, coordinated through an NIH Cancer Moonshot network, independently confirmed that harmonization across sites and techniques is achievable through deliberate coordination. The realistic goal of cross-site validation is not perfect agreement between laboratories, but a quantified, documented benchmark for how much variability is acceptable before a result should no longer be trusted. Co-developing a diagnostic alongside its paired therapeutic, rather than after the drug is already approved, shapes both the clinical trial design and the regulatory timeline.

Key takeaways

A discovery-stage assay and a clinical-trial-ready assay are not the same thing; moving between them, "locking down" the assay, is its own distinct development stage.
A six-site study achieved robust concordance not only for cell density but specifically for coexpression and proximity metrics, the harder, spatial-specific version of the reproducibility question.
A separate multi-institutional effort, coordinated through an NIH Cancer Moonshot network, independently confirmed that harmonization across sites and techniques is achievable through deliberate coordination.
The realistic goal of cross-site validation is not perfect agreement between laboratories, but a quantified, documented benchmark for how much variability is acceptable before a result should no longer be trusted.
Co-developing a diagnostic alongside its paired therapeutic, rather than after the drug is already approved, shapes both the clinical trial design and the regulatory timeline.

What a companion diagnostic requires

The regulatory categories and the general validation discipline a companion diagnostic must satisfy are covered in full in Spatial Biomarkers and Companion Diagnostics: The Next Frontier. What that regulatory framing does not fully convey on its own is how much practical engineering work sits between a validated research finding and an assay that can be run reliably outside the lab that discovered it. That gap is the subject of this guide.

Locking down the assay

A spatial biomarker discovered in a research setting is typically optimized iteratively, with a research team free to adjust staining conditions, antibody concentrations, and analysis parameters as they refine the finding. A companion diagnostic assay cannot work this way; every parameter has to be fixed, documented and then left alone, precisely because a validated assay’s performance characteristics only mean something if the assay itself stops changing.

The companion diagnostic requirements literature states why this distinction matters in blunt terms. Research-use-only application of multiplex immunofluorescence in translational research lacks the rigorous controls and standardization needed to support the stringent reproducibility, sensitivity and specificity requirements that same assay faces once it is intended for clinical trial use. A research-stage protocol, however scientifically sound the underlying biology, is not automatically a clinical-grade assay, and treating "locking down the assay" as its own distinct development stage, rather than an afterthought once the science looks solid, is what prevents that gap from becoming a costly late-stage surprise.

Analytical validation

Once an assay is locked down, analytical validation asks a specific question: does this exact, fixed protocol produce consistent, accurate results, and does it keep doing so across the range of conditions a real clinical laboratory will actually encounter.

For a spatial biomarker specifically, that validation has to cover more ground than a conventional single-marker assay, because a spatial signature is defined by more than one kind of measurement at once. The clearest evidence that this is achievable, not merely theoretically desirable, comes from a coordinated multi-site study built specifically to test it.


Proximity, not just density, held up across six independent sites The Multi-institutional TSA-amplified Multiplexed Immunofluorescence Reproducibility Evaluation, known as MITRE, had six sites perform an automated 6-plex, 7-color assay built around the PD-1/PD-L1 axis, measuring PD-1, PD-L1, CD8, CD68, FoxP3, cytokeratin, and a nuclear counterstain. After staining parameters for each antibody were optimized individually, the combined panel was applied to serial sections from tonsil, breast carcinoma, and non-small cell lung cancer tissue microarrays, using a shared automated staining platform, a shared imaging platform, and locked-down analysis algorithms across all six sites. The result directly answers the harder version of the reproducibility question. Inter-site and intra-site concordance was demonstrated not only for measures like density of specific immune cell subsets, but specifically for coexpression metrics, such as the percentage of PD-L1 expressed on immune cells, and for proximity, the physical distance between PD-1 and PD-L1. That is a spatial-specific metric, not a conventional abundance measurement, holding up reliably across six independent laboratories, which is precisely the kind of evidence a spatial CDx program needs before assuming its own proximity or interaction metric will generalize beyond the lab that discovered it.

Proximity, not just density, held up across six independent sites

The Multi-institutional TSA-amplified Multiplexed Immunofluorescence Reproducibility Evaluation, known as MITRE, had six sites perform an automated 6-plex, 7-color assay built around the PD-1/PD-L1 axis, measuring PD-1, PD-L1, CD8, CD68, FoxP3, cytokeratin, and a nuclear counterstain. After staining parameters for each antibody were optimized individually, the combined panel was applied to serial sections from tonsil, breast carcinoma, and non-small cell lung cancer tissue microarrays, using a shared automated staining platform, a shared imaging platform, and locked-down analysis algorithms across all six sites.

The result directly answers the harder version of the reproducibility question. Inter-site and intra-site concordance was demonstrated not only for measures like density of specific immune cell subsets, but specifically for coexpression metrics, such as the percentage of PD-L1 expressed on immune cells, and for proximity, the physical distance between PD-1 and PD-L1. That is a spatial-specific metric, not a conventional abundance measurement, holding up reliably across six independent laboratories, which is precisely the kind of evidence a spatial CDx program needs before assuming its own proximity or interaction metric will generalize beyond the lab that discovered it.

A second, independent effort reinforces the same conclusion using a different underlying technique. The NCI-designated Cancer Immune Monitoring and Analysis Centers, established as part of the NIH’s Cancer Moonshot Initiative specifically to provide standardized biomarker assays for NIH-sponsored clinical trials, harmonized antibody clones used in a 5-plex assay across multiple sites and found strong agreement in immune cell densities, comparing results generated from a chromogenic multiplex technique against a spectral-unmixing multiplex immunofluorescence approach. Two independent, multi-institutional efforts, using different techniques, both achieving meaningful cross-site concordance, is stronger evidence that harmonization is a solvable engineering problem than either result would be alone.

Clinical validation and co-development

Analytical validation confirms an assay measures what it claims to measure, consistently. Clinical validation confirms that measurement actually predicts the outcome a physician needs it to predict, in the specific patient population and clinical context the diagnostic will be used in.

For a companion diagnostic specifically, this stage is rarely conducted independently of the paired therapeutic’s own clinical trial program. Co-development, running the diagnostic’s clinical validation alongside the drug’s efficacy trial rather than after it, shapes both programs simultaneously: the trial’s patient selection criteria may depend on the diagnostic’s output, and the diagnostic’s clinical validation depends on having enough trial patients, with enough outcome diversity, to establish that its readout actually predicts response. That interdependence is precisely why the sheet’s own framing treats this as a co-development question rather than two separate, sequential validation exercises.

Standardization and reproducibility hurdles

The MITRE and Cancer Moonshot harmonization results above are genuinely encouraging, and they should not be read as evidence that standardization eliminates variability between sites entirely. That is not what these efforts were designed to show, and overstating what they demonstrate would be a mistake a development program could pay for later.

A 2025 review of best practices for multiplex image analysis and data sharing states this precisely: the expectation of these harmonization processes is not to achieve perfect agreement between institutions, but rather to quantify inter-institutional variability following coordinated harmonization efforts. The resulting benchmark, however much residual variability remains after locking down the assay and coordinating across sites, becomes the standard against which a real clinical sample’s result is interpreted, rather than an assumption that the number reported will be identical no matter which site ran the test.

Three practical hurdles recur across the harmonization efforts described above, worth planning for explicitly rather than discovering during a real multi-site rollout.

Antibody lot-to-lot variation. A locked-down protocol still depends on a physical reagent that varies somewhat between manufacturing lots, and a harmonization effort has to characterize that variation rather than assume a fixed protocol guarantees a fixed reagent.
Instrument and imaging platform differences. The harmonization efforts described above achieved concordance in part by standardizing on a shared imaging platform across sites; a program planning to deploy across laboratories using different instruments should expect additional harmonization work beyond what a single-platform study demonstrates.
Image analysis algorithm portability. An algorithm validated and locked down on one site’s data does not automatically perform identically when applied to images generated by a different site’s staining and imaging workflow, which is exactly why both harmonization efforts treated the analysis algorithm as something to lock down and test explicitly, not an afterthought to the wet-lab protocol.

Broader questions of how a validated, standardized spatial assay integrates into a regulated laboratory’s day-to-day compliance obligations, including CLIA requirements across the pre-analytic, analytic, and post-analytic testing phases, are addressed by our colleagues at Lab Manager in CLIA Compliance for Pre-Analytic, Analytic, and Post-Analytic Testing Phases. The clinical and translational workflow questions this development path ultimately feeds into are covered in Translating Spatial Biology Into the Clinic: Digital Pathology and Beyond.

For the regulatory categories and general validation discipline this development path assumes, see Spatial Biomarkers and Companion Diagnostics: The Next Frontier, and for where biomarker development sits within the full spatial biology pipeline, Spatial Biology in Drug Discovery: From Target Discovery to Translational Medicine.

This article was produced under Drug Discovery News’s AI editorial policies.