AI and the Acceleration of Drug Development

Like it or not, artificial intelligence has likely impacted your daily life in some way. Whether you use ChatGPT to organize your day or have begun to integrate AI-driven tools in a clinical setting, it is clear that the era of AI in medicine has arrived. One potential revolutionary impact could be the use of AI to accelerate drug development. The drug development life cycle—long known for its slow pace and a daunting 75% failure rate in Phase 2—is being reshaped with the help of AI to accelerate timelines and improve clinical success.¹ 

Reinventing Drug Development with AI-Driven Target Discovery

Take Novartis, which recently teamed up with the AI‑enabled biotech Relation to accelerate the discovery of new immuno‑dermatology targets.^1,2 In traditional pharma, you often don’t find out you picked the wrong target until the product enters clinical trials—after spending several years and several millions of dollars. Instead of this traditional trial and error model, companies are now looking earlier in the value chain at data-driven precision models that can improve their target selection and reduce their failure rate. By using a “functional cell atlas,” Relation aims to simulate that failure in the computer or in the lab first. By utilizing AI to mine through massive datasets of patient-derived multi-omics, the companies hope to confidently identify causal genes that can be targeted successfully in clinical trials. 

Similarly, another AI biotech, Chai Discovery, promises models that use “computer-aided design” of new antibodies to tackle traditionally challenging targets.³ Instead of physically testing millions of physical antibody candidates, the company believes that their technology will allow researchers to skip directly from in silico design to functional validation, thereby reducing costs and development timelines. In a company preprint, the team reported that their platform can design full‑length monoclonal antibodies with an 86% likelihood of meeting therapeutic developability criteria and with atomic‑level structural fidelity.⁴ 

Eli Lilly is partnering with Insilico Medicine for end-to-end coverage of pharmaceutical research and development through Insilico’s Pharma.AI platform.⁵ Insilico touts an average 12–18-month turnaround time from project initiation to preclinical candidate nomination, stating that it has already nominated 20 preclinical candidates between 2021 to 2024. The company’s lead computer-designed drug for idiopathic pulmonary fibrosis (IPF) is advancing to pivotal trials following positive Phase 2a results. Lilly is also partnering with Nvidia to build a supercomputer-powered “AI factory” designed to train models on internal experimental data and generate actionable research insights.⁶

How AI Is Reimagining Ophthalmic Drug Discovery and Early Disease Detection

In ophthalmology, we have several budding models, especially those that unlock the full potential of daily imaging obtained in our clinics. For example, Moorfields Eye Hospital’s INSIGHT project has curated a massive repository of over 35 million ophthalmic images. Their dataset spans conditions from wet age-related macular degeneration to diabetic macular edema, and they are all seamlessly linked to de-identified clinical records. They are partnering with insitro, a machine learning-enabled drug discovery company. The companies are working together to build a foundation model that can detect subtle, latent signals of neurodegenerative diseases like dementia years before symptoms appear through the identification of novel biomarkers. Of note, insitro is also utilizing Lilly’s TuneLab, a machine larging platform for molecule discovery research. It allows early-stage biotechs to access Lilly’s models in exchange for training data used for further model creation.

In a similar push to leverage AI for data-driven discovery, a consortium of academic and nonprofit institutions launched FAIRhub, an open source platform for sharing and accessing FAIR and AI-ready datasets. The flagship project is a dataset that can enable AI and machine learning to provide critical insights into type 2 diabetes mellitus, including salutogenic pathways to return to health. The AI-READI project aims to collect cross-sectional data on 4,000 people with longitudinal data on 10% of the study cohort across the U.S. They have already released a massive, publicly available, 3.87 TB dataset comprising multimodal data (including clinical records, retinal imaging, environmental sensors and wearables) from 2,280 diverse participants ranging from healthy individuals to those with insulin-dependent Type 2 diabetes. Ultimately, the dataset seeks to fill critical gaps in current research by providing the robust, inclusive data needed to capture the disease’s intricate, multi‑organ dynamics.

Challenges in AI Drug Development

While there has been rapid development in this area, AI in drug development still faces numerous challenges for efficient deployment. For example, the “black box” nature of these models can make it difficult to decipher the logic behind an AI’s decisions.² Furthermore, limited transparency in training protocols can make it hard to validate their clinical utility. This presents a challenge for regulatory agencies vetting the qualifications of an AI technology in drug development.² 

This may be why the FDA has only recently qualified its first AI-powered drug development tool. They cleared a tool to assist in the evaluation of liver biopsies for metabolic dysfunction-associated steatohepatitis (PathAI’s AIM-MASH).⁷ Designed to mitigate the high variability and subjectivity in manual pathologist scoring, the tool analyzes digital tissue images to provide standardized assessments of inflammation, steatosis and fibrosis. 

To accelerate the field responsibly, the FDA has committed to establishing clear guidance on assessing risk associated with AI across various stages of drug development; defining standards for data quality, and creating policies and incentives to support multidisciplinary initiatives across public and private institutions and across key stakeholders in the space.² 

Conclusion

Transitioning from a traditional ”trial-and-error” model in the pharmaceutical pipeline, to a generative AI-driven discovery model, would represent a paradigm shift in our industry. By leveraging foundation models and multi-omic datasets, researchers may be able to simulate failures in silico, design antibodies with atomic accuracy and nominate preclinical candidates in a fraction of the traditional time. Ophthalmology can also become a vanguard of this movement by transforming routine clinical imaging into a treasure trove of biomarkers for ocular and systemic health. 

Even with its promise, the journey toward “AI‑aided” drug development is anything but straightforward. To realize its full potential, the industry and regulatory agencies must overcome the many challenges that currently obscure AI decision-making. As regulatory clarity catches up with computational speed, the next decade of drug development could be defined by human ingenuity and machine intelligence working in tandem to bring life-saving therapies to patients faster than ever before.

References

1. Waldron, J. (2025, December 9). Novartis pens $1.7B atopic disease pact with British biotech. Fierce Biotech. Retrieved from https://www.fiercebiotech.com/biotech/novartis-pens-17b-dermatology-pact-ai-enabled-british-biotech.
2. Poddar, A., Samson, M., Innes, G. K., Liu, Q., Saha, A., Hanger, M., Franzetti, K., ElZarrad, M. K., & Fakhouri, T. H. (2025, November 24). Leveraging Artificial Intelligence in Drug and Biological Product Development: An FDA and Clinical Trial Transformation Initiative Workshop Report. NEJM AI.
3. Incorvaia, D. (2025, December 15). Chai infuses AI drug discovery efforts with $130M series B. Fierce Biotech. Retrieved from https://www.fiercebiotech.com/biotech/chai-infuses-ai-drug-discovery-efforts-130m-series-b.
4. Chai Discovery Team, Boitreaud, J., Chen, R., Dent, J., Fairweather, L., Geisz, D., Greenig, M., Boyd, N., Jain, J., Johnston, B., McPartlon, M., Meier, J., Patil, N., Qiao, Z., Rollins, N., Vicas, N., Wollenhaupt, P., Wu, K., & Yeung, A. (2025, December 1). Drug-like antibody design against challenging targets with atomic precision. bioRxiv. https://www.biorxiv.org/content/10.1101/2025.11.29.691346v1
5. Masson, G. (2025, November 10). Lilly continues AI push, inking $100M-plus research pact with Insilico. Fierce Biotech. Retrieved from https://www.fiercebiotech.com/biotech/lilly-continues-ai-push-inking-100m-plus-research-pact-insilico
6. Incorvaia, D. (2025, October 28). Eli Lilly and Nvidia set to build pharma’s biggest supercomputer. Fierce Biotech. Retrieved from https://www.fiercebiotech.com/biotech/eli-lilly-and-nvidia-set-build-pharmas-largest-ever-supercomputer.
7. U.S. Food and Drug Administration. (2025, December 8). FDA qualifies first AI drug development tool, will be used in MASH clinical trials. Retrieved from https://www.fda.gov/drugs/drug-safety-and-availability/fda-qualifies-first-ai-drug-development-tool-will-be-used-mash-clinical-trials.

Load More