FAQ: What You Need to Know About AI and Analytical Innovation in Biopharma Development

What is an AI protein design model, and how is it being applied to early drug discovery?

Protein design models use AI to learn from the sequences produced by biological evolution and translate that information into the ability to predict protein structure and design novel protein binders. For example, Biohub, a nonprofit biomedical research organization associated with the Chan Zuckerberg Initiative, released an open-source suite of protein biology models in May 2026. These models are built on fourth-generation evolutionary scale modeling (ESM4).¹

The suite, which comprises ESMFold2, the Evolutionary Scale Modeling Cambrian (ESMC) protein language model, and the ESM Atlas covering 6.8 billion proteins, was described by the organization as a world model of protein biology capable of mapping proteins across the tree of life, predicting structures, and designing new binders.¹ In early laboratory experiments, researchers used the models to design protein binders targeting cancer and immune-related proteins, with the binders reportedly reactivating immune cells in vitro.² Translation to clinical candidates still requires evaluation of stability, specificity, manufacturability, and in vivo activity.

How are large biopharmaceutical companies integrating AI into research and development decision-making?

Rather than deploying AI for isolated tasks, companies are increasingly seeking to embed continuous-learning platforms throughout the discovery and development lifecycle. Incyte and Edison Scientific, for example, announced a strategic collaboration in May 2026 to integrate Edison's Kosmos AI platform, described as an AI scientist, across Incyte's research workflows.³

The initial deployment is focused on target discovery, target validation, and translational biology, with the system designed to analyze experimental, clinical, and biomarker data and generate predictive models of therapeutic performance. According to the companies, the Kosmos platform is intended to enable continuous learning from translational and clinical data and provide real-time synthesis of evidence.³ The collaboration reflects a broader industry trend in which AI is used not merely to analyze existing data, but to function as a learning system that compounds knowledge with each successive experiment or clinical readout.⁴

What role does mechanistic modeling play in advancing chimeric antigen receptor T cell (CAR-T) therapies?

Chimeric antigen receptor T cell (CAR-T) therapies present unique pharmacokinetic (PK) and pharmacodynamic challenges because they function as living drugs. After infusion, CAR T cells can proliferate rapidly in response to tumor antigens, which leads to drug concentrations that may increase by orders of magnitude beyond the administered dose.⁵

Such behavior makes traditional PK modeling insufficient. Quantitative systems pharmacology (QSP) approaches mechanistically characterize immune activation, tumor response, and CAR T-cell persistence simultaneously, enabling more rigorous predictions of both efficacy and toxicity.⁵

"Pharmacologic QSP is a mechanistic framework which makes an effort to understand and quantify in quantitative, in mechanistic, predictable terms the processes that take place, from tissue distribution, target engagement, activation, proliferation, differentiation, eventual elimination of the cells, and how it is all related to the dose, to the antigen, density, the location of the cells," says Armin Sepp, PhD, senior principal scientist and head of Biologics Modeling at Certara UK.⁵

Published mechanistic modeling frameworks for CAR-T therapy have demonstrated the ability to capture cellular kinetics and pharmacodynamic outputs, including the impact of tumor antigen regulation and patient-specific factors on treatment outcomes.⁶

How do multi-attribute methods (MAMs) improve critical quality attribute monitoring for bioconjugates?

Antibody-drug conjugates (ADCs) and other bioconjugates introduce analytical complexity beyond conventional monoclonal antibodies, including variable conjugation sites, drug-to-antibody ratio heterogeneity, and novel degradation pathways.⁷ Multi-attribute methods (MAMs) based on liquid chromatography–mass spectrometry peptide mapping address this complexity by quantifying multiple critical quality attributes, including post-translational modifications, sequence variants, and conjugation-site occupancy, within a single, integrated assay format.⁸

This consolidated approach enables faster turnaround and more consistent results compared with parallel orthogonal methods, and is particularly valuable in stability and comparability studies where detecting subtle changes across manufacturing batches is essential.⁸

“Platforms like multi attribute methods are no longer exploratory tools. They are basically the foundational enablers of a faster development, better decision, and regulatory-ready control strategies for the end user," says Ganesh Bala, PhD, product specialist, Agilent Technologies.⁸ Published workflows for MAMs have validated this approach across both biotherapeutic process development and quality control applications.⁹

How is AI being used to manage supply chain complexity for advanced therapy modalities?

Radiopharmaceuticals and other short-lived advanced therapy modalities impose supply chain requirements that differ fundamentally from those of conventional biologics. Such modalities must be produced, tested, shipped, and administered within narrow time windows, often across international borders with varying regulatory requirements. AI-enabled logistics systems are said to be deployed to support real-time coordination, route optimization, and predictive management of delivery constraints for radiopharmaceutical programs.¹⁰

Growth in the global radiopharmaceutical market,which is projected to reach more than $26 billion by 2031 from approximately $9 billion in 2023, highlights the urgency and need for scalable logistics solutions.¹⁰

References

1. Biohub. Biohub releases a world model of protein biology. Published May 27, 2026. Accessed June 4, 2026. https://biohub.org/news/world-model-of-protein-biology/
2. Mirasol F. Biohub open-source AI model targets protein design for drug discovery. BioPharm International. Published May 28, 2026. Accessed June 4, 2026. https://www.biopharminternational.com/view/biohub-open-source-ai-model-targets-protein-design-for-drug-discovery
3. Incyte. Incyte and Edison Scientific announce strategic collaboration to employ the Kosmos AI platform for research and development. Published May 19, 2026. Accessed June 4, 2026. https://investor.incyte.com/news-releases/news-release-details/incyte-and-edison-scientific-announce-strategic-collaboration
4. Schoenthaler E. Incyte partners with Edison Scientific to integrate AI across drug discovery and translational research. BioPharm International. Published May 19, 2026. Accessed June 4, 2026. https://www.biopharminternational.com/view/incyte-partners-with-edison-scientific-to-integrate-ai-across-drug-discovery-and-translational-research
5. Mirasol F, Sepp A. How mechanistic modeling advances CAR T-cell therapy development. BioPharm International. Published May 13, 2026. Accessed June 4, 2026. https://www.biopharminternational.com/view/how-mechanistic-modeling-advances-car-t-cell-therapy-development
6. Minucci S, Gruver S, Subramanian K, Renardy M. A multi-scale semi-mechanistic CK/PD model for CAR T-cell therapy. Front Syst Biol. 2024;4:1380018. doi:10.3389/fsysb.2024.1380018
7. Zhou X, Han Y, Fang Y, et al. Antibody-drug conjugates: current challenges and innovative solutions for precision cancer therapy. Med. 2025;6(10):100849. doi:10.1016/j.medj.2025.100849
8. Mirasol F, Bala G. A Q&A with Dr. Ganesh Bala on LC–MS and multi-attribute methods for bioconjugate CQA monitoring. BioPharm International. Published June 4, 2026. Accessed June 4, 2026. https://www.biopharminternational.com/view/q-a-ganesh-bala-lc-ms-multi-attribute-methods-bioconjugate-cqa-monitoring
9. Millán-Martín S, Jakes C, Carillo S, Bones J. Multi-attribute method (MAM) analytical workflow for biotherapeutic protein characterization from process development to QC. Curr Protoc. 2023;3(11):e927. doi:10.1002/cpz1.927
10. Schoenthaler E, Hogenboom M, Zobel A. Supply chain execution in radiopharma: the industry's next critical challenge. BioPharm International. Published May 26, 2026. Accessed June 4, 2026. https://www.biopharminternational.com/view/supply-chain-execution-in-radiopharma-the-industry-s-next-critical-challenge