Target antigen expression has long served as a proxy for predicting how patients will respond to antibody-drug conjugates (ADCs). A recent collaboration between Nucleai, a company developing AI-based tissue analysis tools, and Gilead Sciences suggests that measure alone doesn't tell the full story.
The two companies analyzed a large set of hematoxylin and eosin (H&E) and immunohistochemistry (IHC) whole-slide images from several clinical studies spanning multiple oncology indications. Using AI-driven tissue analytics alongside clinical outcomes, the analysis generated candidate spatial biomarkers the companies plan to detail in future scientific presentations and publications.
The collaboration reflects a broader trend among drug developers working on ADCs, a class of therapies that link cytotoxic payloads to antibodies targeting specific antigens on tumor cells. As more ADCs move through clinical development, companies have increasingly reported cases where target expression levels don't cleanly predict which patients will respond, prompting a search for additional biomarkers that can better stratify patients.
Why target expression isn't enough
According to Ken Bloom, Head of Pathology at Nucleai, target expression remains associated with clinical outcomes, but spatial characteristics of the tumor and its microenvironment appear to carry comparable predictive value. Features drawn from both IHC and H&E, including target spatial organization, tumor architecture, and the composition of the tumor microenvironment, provided complementary information about treatment response beyond target levels alone.
Bloom told DDN that the findings point toward a shift from single-marker quantification toward more comprehensive spatial characterization when developing biomarkers for ADC therapies. Rather than treating target expression as a standalone readout, the approach folds it into a broader picture of how that target is distributed across the tumor and what surrounds it, including immune cells, stromal architecture, and other features of the tumor microenvironment that conventional scoring methods don't capture.
What scale changes
Conventional pathology typically involves a pathologist scoring one or two markers at a time in a representative slide region, an approach Bloom described as semi-quantitative and subject to reader-to-reader variability. Analyzing whole-slide images at scale, he said, converts that qualitative read into a standardized measurement applied consistently across every cell on a slide, enabling comparisons across studies and sites.
It also allows researchers to quantify spatial features, such as cell distances, densities, and neighborhood patterns, that are difficult for a human reviewer to assess reliably across large numbers of slides. Rather than a pathologist estimating a percentage of positive cells in a chosen field, the platform measures every cell across the full slide, then tests many features and thresholds systematically against clinical outcomes to identify which most reliably separate responders from non-responders.
Because H&E slides already exist for most patients enrolled in completed trials, Bloom said archived pathology material can be mined retrospectively rather than requiring new tissue collection for future biomarker studies. That distinction matters for drug developers working with limited tissue samples from completed trials: instead of spending finite biopsy material on a new assay, existing slide archives become a resource for discovering and testing new biomarker hypotheses after the fact.
What it means for drug developers
For teams running or planning ADC programs, the work points toward biomarker strategies that look beyond target expression scoring alone. Standardizing spatial analysis across large sample sets, rather than relying on manual reads that vary between pathologists and sites, could make it easier to compare biomarker performance across trials and institutions.
Nucleai said the broader Tissue Intelligence platform is intended to support biomarker development from preclinical research through late-stage clinical trials, and that the collaboration with Gilead reflects a wider pattern of pharmaceutical companies integrating tissue architecture data into translational and biomarker programs.
“Precision oncology is entering a new phase, where understanding tissue architecture is becoming just as important as understanding molecular alterations,” Avi Veidman, CEO of Nucleai, said in the press release. “Tissue Intelligence is becoming a foundational capability for precision medicine, helping identify the patients most likely to benefit while enabling pharmaceutical companies to translate tissue biology into reproducible biomarkers that improve the speed and success of oncology drug development.”
