Digital pathology and AI are rapidly reshaping how oncology teams discover, validate and deploy tissue-based biomarkers. In this thought leadership interview, Javier Perez, Senior Director of Precision Medicine Diagnostics and Digital Pathology Strategy in Oncology at Regeneron, shares his perspective on where these technologies are creating value today and what will be required to move them from translational research into regulated diagnostic workflows.
According to Perez, the most immediate value of digital pathology lies in biomarker discovery and biomarker development. While diagnostic development is advancing, he believes the field is still in its early stages, with significant progress expected over the next few years as AI applications become more deeply embedded in therapeutic co-development.
One of the most promising opportunities is the ability to extract new biological insights from routine tissue images. Perez points to the potential of AI models trained on H&E-stained slides to identify patterns associated with recurrence, treatment response or patient outcomes, even without additional IHC staining. This could open new routes for discovering candidate biomarkers that are later validated and translated into clinically actionable tools.
For a digital pathology biomarker to influence clinical development decisions, robustness is essential. Perez highlights the quantitative nature of computational pathology as a major strength, particularly in improving accuracy and precision in IHC interpretation. In areas where pathologist reads can be variable or equivocal, AI-enabled approaches may help standardise assessment and uncover biomarkers that are not readily visible to the human eye.
However, Perez is clear that value must be demonstrated against existing benchmarks. AI pathology models need to show that they add predictive value, identify additional relevant patients, or expand on what current biomarker strategies can achieve. At a minimum, they should be non-inferior to established workflows, while also justifying the added operational complexity, cost, quality control and implementation burden.
In oncology trials, Perez sees near-term potential in patient stratification and enrichment, but only once biomarkers have been sufficiently validated. Candidate biomarkers discovered through AI must be biologically understood, operationally deployable and clinically meaningful before they can be used prospectively to reduce development risk.
Building confidence in AI-enabled pathology will require collaboration across pharma, diagnostic partners, pathologists, technology developers and regulators. Perez emphasises the importance of clinical validation, retrospective evidence and cross-sector consortia in advancing the field. While digital pathology is already being used in clinical workflows to manage slides, share data and support efficiency, broader diagnostic co-development with oncology therapies will depend on clear proof that these tools perform reliably and add meaningful value.
Looking ahead five years, Perez expects computational pathology to become a routine component of oncology research and development. He anticipates its use in retrospective and translational analyses, exploratory clinical trial endpoints, patient stratification strategies and, potentially, approved companion diagnostics based on computational pathology.
For precision oncology, the message is clear: digital pathology and AI are no longer simply research tools. With the right validation, workflow integration and regulatory confidence, they have the potential to become central to how oncology biomarkers are discovered, developed and deployed.
