In Industry Insights, members of ASPET’s Industry Science Committee discuss the intersection of pharmacology and industry, private sector highlights, and how the industry and membership can support each other.
Where Science Meets Silicon: AI’s Growing Impact on Drug Discovery
By Hiranmayee Kandala, PhD (Principal Scientist, Amgen) and Abdalla Wedn (Pharmacological PhD Candidate, University of Pittsburgh)
Artificial intelligence is becoming a practical force across drug development, supporting scientific decision making from early discovery through clinical design. Today, AI plays an active role in target identification, hit and lead optimization, PK/PD modeling, and clinical trial planning. Machine-learning models analyze large genomic and chemical datasets to uncover biological relationships, propose new molecular structures, predict drug behavior, and highlight potential risks far earlier than traditional workflows.
Pharmacologists, chemists, and biologists now work closely with these tools as part of their everyday research. Modern systems range from machine-learning models that iteratively refine predictions as new data accumulates to generative AI algorithms capable of designing novel molecules or simulating biological outcomes. These approaches provide rapid, data driven insights that help teams prioritize experiments, refine hypotheses, interpret complex datasets, and move through design cycles efficiently. Scientists remain central to every decision, using computational insights to validate ideas and guide programs toward promising directions.
Applications of AI span discovery through development and commercialization. In lead discovery, AI-supported biotransformation prediction and metabolite profiling are essential for understanding structure activity relationships. Tools such as Mass MetaSite, Metabolite Profiler, and Compound Discoverer help identify metabolic soft spots and interpret mass spectrometry data. In hit and lead design, generative chemistry models tailor molecules with improved potency, selectivity, and ADME properties. Machine learning also enhances PK/PD modeling by providing early estimates of human exposure and clearance, and clinical development has adopted AI-enabled patient stratification, trial site selection, and continuous safety monitoring.
While AI offers transformative capabilities, it cannot overcome the biological and regulatory constraints that shape drug development timelines. It can design molecules and accelerate decision making, but it cannot replace toxicology studies, shorten the natural course of disease in trials, or alter the biological constants that govern experiments. Its value lies in compressing discovery phases when paired with deep scientific expertise, potentially reducing timelines by a few years while supporting rigorous testing required to ensure drug safety and efficacy.
Recognizing both AI’s promise and its limitations, the United States Food and Drug Administration is cautiously integrating AI into regulatory processes. In January 2025, the agency released draft guidance outlining a risk-based credibility assessment framework for AI models used across development, from nonclinical studies to manufacturing. In June 2025, the FDA launched Elsa, a generative AI tool designed to accelerate internal review processes, with tasks that once took days now completed in minutes. Elsa is already being deployed to streamline clinical protocol reviews and support safety assessments.
Taken together, AI is not replacing drug discovery, it is rewiring it. The power of AI lies in amplifying human expertise, accelerating bottlenecked steps, and revealing insights hidden within biological complexity. As data quality and regulatory frameworks mature, AI is expected to evolve from a helpful companion to an indispensable foundation of modern drug development. Teams that combine scientific intuition with computational intelligence will shape the next generation of breakthroughs. AI is not the future of drug development; it is the foundation being built right now.