Turn early tissue signal into

evidence a program
can act on.

Nucleai Insight quantifies what is actually on the slide, then tests it against clinical outcomes. It answers key questions translational research keeps returning to: what is in this tissue, why did the drug work, and in whom. And the analysis can run inside your own environment when your data cannot leave it.

One offering, two questions

Translational research asks the same tissue two different things. Early in a program, before there is outcome data to work with, the question is what is actually present: which features can be measured consistently enough to build on. Later, once a cohort has matured and clinical outcomes are in hand, the question becomes why the drug worked, and in whom.

Nucleai Insight answers both, through two focused engagements built on the same underlying platform. Analyze quantifies the tissue itself. Profile & Predict tests those features against a program's clinical outcomes. Which one a sponsor needs depends on where their data stands, and many programs move through both.

NUCLEAI INSIGHT · Analyze

Analyze: What is actually in this tissue?

Most sponsors can already count cells with commercial tools. The problem is precision: standard measurement is coarse, binned rather than continuous, inconsistent across readers and sites, and blind to the spatial relationships that newer targets depend on. Analyze is built for programs where that measurement gap is the actual bottleneck.

Nucleai extracts continuous, reproducible feature scores across the full expression spectrum. Real programs mix image modalities: H&E on the entire cohort, IHC on a subset, and sometimes multiplex imaging. Nucleai runs all modalities through one analytical framework so features stay comparable across those slices instead of coming back from three separate vendors. The result is a structured feature dataset plus an interpreted scientific report.

Evidence

In a multi-site validation with Merck KGaA presented at USCAP 2023, a deep learning PD-L1 TPS scoring model was trained across five laboratories, three scanners, and two antibody clones on roughly 100,000 pathologist-annotated cells. It was then validated against two independent pathologists on an unseen 107-slide cohort at an R² of 0.91 against pathologist consensus. In a separate study presented at SITC 2023, pairing low-plex multiplex imaging with same-slide H&E reclassified six million previously marker-negative cells and identified tertiary lymphoid structures at 88% accuracy.

Two ways to start

Both of the following are scoped engagements inside Analyze, built to answer the most common initial questions.

NUCLEAI INSIGHT · Profile & Predict

Profile & Predict: why did the drug work, and in whom?

The predictive power of single markers is limited, and testing that requires outcome data that often cannot leave the sponsor's building. Left unconstrained, high-dimensional analysis of tissue can surface more findings than any team can interpret, with a well-founded risk that few of them replicate.

Profile & Predict tests spatial features and thresholds systematically against a program's clinical and outcomes data, and can run inside the sponsor's own environment. Findings are built to survive a second cohort, with held-out validation and cross-cohort replication as the design standard, and are interpreted through Kaplan-Meier plots and threshold analyses that a program committee can act on.

Evidence

In a peer-reviewed collaboration with Adlai Nortye published in Cancers (2026), spatial features read from randomized Phase 2 H&E were associated with overall survival benefit.

In a mechanism-linked study with Debiopharm presented at ENA 2022, spatial features were tied to response in an anti-CD37 antibody-drug conjugate program.

And in a collaboration with Genmab presented at SITC 2022, adding transcriptomic data to H&E and IHC features raised model accuracy in distinguishing outcomes from an AUC of 0.65 to 0.76 on the same cohort and task, alongside a progression-free survival hazard ratio of 0.48.

Proven in translational research

Demonstrating our Nucleai Insight Profile & Predict offering, an ongoing collaboration supporting Gilead's global antibody-drug conjugate programs has analyzed thousands of H&E and IHC whole-slide images across multiple oncology indications, connecting tissue biology to clinical outcomes at scale.

Why translational research starts here

Most tissue analysis stops at percent positivity and an H-score, a single number that hides biological complexity. Nucleai transforms tissue images into quantitative, spatially resolved biology, with tissue analysis across imaging modalities, systematic linkage to clinical outcomes, and findings that trace to the underlying tissue, interpreted by expert pathologists and scientists—not delivered as raw output.