Artificial intelligence (AI) is playing an expanding role in advancing precision medical therapies. The inherent complexity and variability of advanced therapy medicinal products (ATMPs) demand innovative strategies for such activities as potency evaluation, real-time process surveillance, and stability assessment (1). AI technologies can support these domains while operating within increasingly rigorous regulatory frameworks and the rising importance of global data protection requirements—emphasizing secure data management and patient privacy throughout ATMP development and analysis. At the same time, the adoption of AI requires careful attention to explainability and transparency. To ensure compliance and build trust in AI-generated insights, techniques such as SHapley Additive exPlanations (SHAP) values and physics-informed neural networks (PINNS) are proving valuable.

To begin with, high-dimensional data are becoming essential to many aspects of precision medicine, including many ATMP. These data provide the multidimensional molecular and clinical information required to decipher complex disease mechanisms and individual variability in precision medicines (Table 1). AI demonstrates exceptional capability in extracting patterns, relationships, and predictive insights from such datasets (2).

AI is proving to be powerful in advancing many aspects of ATMP establishment, including product discovery, regulatory compliance, and manufacturing operations (3, 4). The modern therapeutic paradigms variously labeled precision or personalized medicine represent an interconnected and evolving landscape. While sharing related goals, the different therapeutic strategies have distinct aims and implementations designed to make treatments more individualized, effective, and patient-centered. These specialized therapeutic models are informed by prior screening that reflects a patient’s baseline genetic, cellular, or metabolic profile, or their current treatment response. They can be broadly grouped into two approaches. The first involves creating a unique therapeutic agent, dosage, or regimen designed specifically for an individual patient. The second personalizes treatment by stratifying patients and identifying appropriate candidates for established therapies or delivery routes.

ATMPs encompass gene therapies, somatic cell therapies, and tissue-engineered products. Their design and application rely heavily on biomarker-driven patient selection and genomic profiling to guide candidate identification and/or individualized manufacturing. Autologous cell therapies, such as chimeric antigen receptor T-cell (CAR-T) treatments, represent the most established form of the first personalization model, in which a therapy is generated from a patient’s own cells. Because these cells originate from the individual being treated, such therapies reduce the likelihood of immune rejection.

A clear example of the second personalization approach addresses HER2-positive breast cancer, characterized by overexpression of the HER2 protein. Treatment modalities include monoclonal antibodies that bind HER2 receptors and tyrosine kinase inhibitors that block HER2 signaling. Here, personalization lies in screening: only patients with specific biomarkers are likely to benefit from an established therapy.

AI systems enhance patient stratification by 1) improving the interpretation of predictive genetic markers, such as single nucleotide variants, copy number changes, or structural alterations, and 2) constructing patient similarity networks using multi-omics and clinical profiles through graph neural networks (GNNs), thereby increasing classification accuracy. In this way, AI significantly advances genetic and molecular profiling to detect mutations or variations that shape disease susceptibility and drug response, ultimately guiding tailored therapeutic strategies.

Validated AI algorithms support whole exome sequencing (WES) for identifying genetic conditions, including monogenic diseases and cancer-related mutations. WES generates millions of short DNA reads, which AI tools annotate and analyze against reference genomes. This approach has become vital in diagnosing conditions such as Marfan syndrome, cystic fibrosis, and rare inherited disorders. In parallel, AI is being applied to predict potential off-target effects in DNA-editing therapies, improving safety by reducing unintended modifications.  In pharmacogenomics, AI is enhancing the discovery of genetic factors affecting drug metabolism, thereby informing treatment selection, minimizing adverse drug reactions, and refining dosage optimization.

In highly regulated contexts, deploying complex machine learning models for treatment selection requires interpretability and auditability. Integrating such advances as explainable AI (XAI) methods support this requirement. For instance, interpretable models applied to weighted input samples can clarify which features contributed most to a prediction. Such AI-driven methods are equally powerful in analyzing physicochemical and multi-omic datasets. Compliance with data-protection regulations is ensured through mechanisms such as encryption, access controls, and audit trails. Federated learning frameworks now allow decentralized model training across multiple clinical centers without directly sharing sensitive patient data.

The evolution of regulatory frameworks and best practices is helping integrate AI into advanced therapies, and updates in this arena are frequently communicated through major conferences and congress. For instance, the Bioprocessing Summit Europe, already regarded as a premier European bioprocessing event, now focuses on improving biomanufacturing quality and control of such emerging biologics as ATMPs. Last year they introduced a new session on AI and Process Control featuring such papers as Physics-Informed Artificial Intelligence: A Groundbreaking Technology in the Biopharmaceutical Industry from Ignasi Bofarull-Manzano of RWTH Aachen University. For 2026 they have already secured Jack Prior, Head of Process Monitoring & Data Science and AI strategy at Sanofi who will speak on Advancing Biomanufacturing through Digitalization, Process Modeling and AI (Summit 2026).

The intrinsic complexity and heterogeneity of advanced therapy medicinal products (ATMPs) establish AI’s value in potency determination, real-time process monitoring, and stability assessment. AI-based approaches hold substantial promise in expediting product discovery, enhancing regulatory compliance, and advancing the efficiency of biomanufacturing operations (5).

About the Author

William Whitford

Bill is Founder of Oamaru BioSystems with over 20 years experience in biotechnology product and process development. He now publishes oral papers, print articles, and book chapters on such topics as ATMP process intensification, and AI/ML tools in biomanufacturing. Recently his work has been acknowledged in the 2022 APEX Award for Publication Excellence in the Technical & Technology Writing category and the 2023 ISPE Roger F. Sherwood Article of the Year award. He currently enjoys serving on such committees as the BioProcess International Editorial Advisory Board, and the chair of the 3SMAGNET Industry Advisory Board.