Artificial intelligence is rapidly redefining how cancer biomarkers are discovered, measured, and translated into clinical decisions. Nowhere is this shift more visible than in computational pathology, where AI is enabling precision medicine at a level of consistency and scale that manual scoring cannot match.
Across the industry, collaborations between pharma companies, diagnostics developers, academic medical centres, and AI technology partners are accelerating this transformation. These partnerships are focused on building AI-enabled assays that can robustly identify the right patients for targeted therapies — particularly antibody-drug conjugates (ADCs), where patient selection depends on detecting subtle but clinically meaningful patterns of protein expression.
A leading example in the field is the move toward quantitative, continuous biomarker scoring. Rather than relying on traditional HER2-style ordinal categories, AI models can measure expression on a continuous scale and distinguish patterns that are difficult for humans to reproduce consistently. In the case of TROP2-targeted ADCs, for instance, AI-based methods have been developed to quantify membrane versus cytoplasmic expression and apply defined cutoffs with high reproducibility — effectively creating diagnostics that are only feasible through computational analysis.
Similar collaborative efforts are underway for other emerging targets. Industry teams are co-developing AI-assisted algorithms to score biomarkers such as FGFR2b in gastric cancer and other solid tumour settings, with the goal of standardising interpretation across sites and trial cohorts. At the same time, joint research programmes are expanding AI’s role beyond intensity scoring into spatial biology — using models to quantify features like the “bystander effect” in ADC-treated tumours by mapping how biomarker-positive and biomarker-negative cells interact within the tumour microenvironment.
The result is a new paradigm for biomarker development: pathologists remain essential for biological and clinical context, while AI contributes highly quantitative, reproducible measurement. Together, these cross-sector collaborations are enabling biomarker definitions that were previously impossible to capture reliably.
Computational pathology is no longer experimental — it is quickly becoming core infrastructure for precision oncology. As collaborative AI-based assays move into both clinical trials and routine diagnostic workflows, the trajectory is clear: pathology is becoming more algorithmic, more quantitative, and increasingly central to precision diagnostics and therapeutic development.
