AI Redefines Precision Oncology at ESMO 2026

Oncology has spent two decades expanding its therapeutic targets. KRAS-directed agents, Antibody-drug Conjugates (ADCs) and bispecific antibodies have widened the range of biologically defined patient populations that can benefit from treatment. Yet each new biomarker-defined indication creates a downstream challenge: a targeted therapy can deliver its intended value only if eligible patients are identified quickly, reliably and equitably. Precision oncology is therefore entering a new phase in which the priority extends beyond detecting molecular alterations to integrating histopathology, genomic profiles and clinical data into more informed treatment decisions.

This is where Artificial Intelligence (AI) is gaining relevance, not as a replacement for clinical expertise, but as a bridge between complex data and actionable decisions. By extracting signals from pathology images, sequencing outputs and real-world clinicogenomic records, AI could refine biomarker assessment, estimate recurrence risk and improve patient stratification. At the European Society for Medical Oncology (ESMO) Congress 2026, led by Congress President Fabrice André, with Scientific Co-Chairs Yelena Janjigian and James Larkin, this proposition will be examined through anticipated presentations, including Agendia’s late-breaking research on an AI-driven approach to the MammaPrint 70-gene recurrence signature, AI-enabled prescreening in Non-small Cell Lung Cancer (NSCLC) and AI-based analysis of the tumor microenvironment. The central question is whether these applications can move beyond technical accuracy to demonstrate meaningful clinical utility. Ultimately, their value will depend not simply on how effectively they identify patterns, but on whether those insights improve biomarker identification, guide treatment selection and translate into better patient outcomes.

“The next frontier in precision oncology lies not merely in identifying new therapeutic targets, but in accurately and reliably identifying the patients most likely to benefit from targeting them.”

Beyond Conventional Biomarkers: Why AI Matters Now

Precision oncology has traditionally relied on complementary sources of evidence, including genomic alterations, gene-expression signatures, immunohistochemistry, histopathology and clinical characteristics. These approaches have enabled targeted therapies and biomarker-guided treatment, yet each captures a distinct aspect of tumor biology. As therapeutic development increasingly depends on precise patient selection, the challenge is to integrate these fragmented signals into a more comprehensive understanding of disease.

AI could help bridge this gap by translating complex datasets into clinically relevant insights. Computational pathology can quantify morphological features across whole-slide images and characterize the Tumor Microenvironment (TME), while multimodal models can integrate pathology with genomic, radiological and clinical data. Clinicogenomic analyses can further explore relationships between molecular profiles, treatment exposure and patient outcomes. Together, these approaches could extend biomarker assessment beyond individual molecular alterations toward a more integrated view of tumor biology and treatment response.

Three developments are particularly relevant:

• From visual assessment to computational pathology: AI can systematically quantify tissue morphology and cellular features, potentially improving assessment consistency and identifying patterns that may be difficult to capture through routine visual examination alone.

• From individual biomarkers to multimodal signatures: Integrating histopathology, molecular profiles and clinical characteristics could provide a more comprehensive basis for risk stratification and patient selection than isolated biomarkers.

• From retrospective associations to clinical utility: The critical next step is to establish whether AI-derived insights can improve diagnostic pathways, guide treatment selection, refine risk assessment or support clinical trial enrollment.

These developments do not make established biomarkers or validated assays obsolete. Rather, they position AI as a potential complementary layer of precision oncology, one that could extract additional information from existing data, prioritize patients for further testing and strengthen clinical decision-making. The defining question is whether these computational insights can deliver reproducible, clinically meaningful benefits beyond current standards of care.

Five AI-centred Abstracts to Watch at ESMO 2026

ESMO abstracts are published at the congress, and no objective measure of "most talked about" exists before the meeting opens. This editorial selection highlights AI-focused research publicly announced by participating organizations, prioritizing clinical relevance, breadth across tumor types and implications for biomarker-driven drug development. As the results have not yet been reviewed, no efficacy or performance conclusions are drawn.

 

The Next Frontier: From Single Biomarkers to Multimodal Intelligence

The five studies illustrate how AI could connect information traditionally assessed in isolation, integrating histopathology, molecular profiling, medical imaging and clinical outcomes to provide a more comprehensive view of tumor biology. By identifying relationships between morphological features, genomic alterations and treatment outcomes, multimodal AI could refine patient stratification, support biomarker discovery and generate hypotheses for clinical investigation.

For pharmaceutical companies, these capabilities could inform biomarker development, patient selection and clinical trial design, while diagnostic developers could use AI to complement established assays and prioritize additional testing. However, integrating multiple data sources does not guarantee better predictions. Clinical adoption will depend on demonstrating incremental value over existing approaches, validating performance across diverse populations and establishing that AI-derived insights meaningfully improve clinical decision-making.

Commercial and Clinical Implications: Where Could the Value Emerge?

The commercial and clinical value of AI-enabled precision oncology spans three areas. 

• Companion and complementary diagnostics: AI-derived pathology signatures could support patient prescreening, prioritize molecular testing and complement established assays, provided they demonstrate clinical validity and added value. 

• Drug development and biomarker-led trials: computational pathology and clinicogenomic data could inform patient selection, trial design and biomarker discovery, although retrospective associations alone cannot establish predictive treatment benefit. 

• Clinical workflows and access: automated image analysis could improve assessment consistency and testing efficiency, subject to validation in routine practice.

Companies such as Tempus are expanding their presence in AI-enabled research and clinicogenomic data platforms, reflecting growing industry interest. However, platform capabilities do not automatically establish the clinical utility of individual applications. Commercial adoption will depend on robust validation, integration into clinical workflows, regulatory compliance and demonstrable benefits over existing approaches.

The Barriers That Will Determine Clinical Adoption

Despite its potential, AI-enabled precision oncology faces several challenges that will determine whether these technologies progress beyond research settings.

• Clinical utility: Does using the model improve diagnostic decisions, treatment selection or patient outcomes compared with current practice?

• External validation: Does performance remain consistent across hospitals, scanners, staining protocols and patient populations?

• Bias and representativeness: Do training and validation datasets adequately represent the populations in which the model will be deployed?

• Interpretability and accountability: Can clinicians understand the model's limitations, assess uncertainty and recognize when its output should not guide care?

• Regulatory and economic viability: Can the technology meet applicable regulatory requirements and demonstrate sufficient clinical and economic value to justify adoption?

Concordance with established genomic assays is an important milestone for AI-derived surrogates, but clinical utility requires evidence of comparable prognostic performance and decision-making value. Similarly, predictive biomarkers must demonstrate differences in treatment benefit, not merely prognosis. At ESMO 2026, the most meaningful advances will be those that demonstrate robust validation, reproducibility and measurable value beyond existing clinical approaches.

Conclusion

ESMO 2026 highlights the growing role of AI in advancing precision oncology, from AI-driven digital pathology and biomarker prescreening to the analysis of immune biology and the TME. While these applications could improve patient stratification, streamline diagnostic workflows and support biomarker-led drug development, their long-term significance will depend on robust validation and demonstrable clinical utility. The next frontier is not simply identifying patterns that conventional methods may overlook, but translating those insights into more accurate, actionable treatment decisions. Ultimately, AI will establish its value in oncology by demonstrating that it can complement established biomarkers, improve clinical decision-making and contribute to better patient outcomes.