The past few years have seen AI emerge as one of the most significant advances in drug discovery. From predicting protein structures to designing novel molecules and simulating drug-target interactions in silico, AI is enabling researchers to generate and prioritize biological hypotheses at an unprecedented scale.
And it is easy to see why the excitement has taken hold. Drug discovery is a notoriously expensive and uncertain process, where years of work can end in failure at any stage of development. By helping researchers work more efficiently, AI is increasing the speed, scale, and ambition of discovery.
Key takeaways:
- AI is accelerating the speed and scale of early-stage drug discovery, allowing researchers to generate and prioritize more biological hypotheses than ever before.
- As AI expands what can be explored computationally, the need for high-quality biological validation also increases, making experimental evidence essential.
- The industry is moving toward a Lab-in-the-loop approach, where computational and experimental workflows continuously inform one another to accelerate discovery and improve predictive performance over time.
According to a recent analysis by William Blair,1 AI is already reshaping key stages of the drug discovery pipeline, particularly in early-stage work where biological data is too large and complex to process manually. It is being used to support faster target identification and lead discovery, while helping prioritize candidates for experimental testing.
But reshaping a process is not the same as replacing it. As AI expands the number of hypotheses, targets, and candidate molecules that can be explored, it also increases the need for biological evidence to validate and prioritize those outputs. AI can help researchers decide what to test, but it must be paired with experimental validation to confirm biological relevance. And when it comes to the wet lab, those closest to the technology tend to agree: its importance only grows as AI increases the scale and potential of drug discovery.
AI as a “dry lab” tool
AI is increasingly being leveraged as a “dry lab”, or computational, tool in drug discovery to expand hypothesis generation, support prioritization, and improve the interpretation of complex biological data at scale. Two areas that illustrate this well are target identification and lead discovery.
At the target identification stage, machine learning models are used to mine genomic, proteomic, and clinical datasets at a scale that would be impossible to achieve manually. These models help prioritize potential disease-relevant targets, predict druggability, and flag associations that might otherwise take years to uncover through conventional approaches.
In lead discovery, generative models are now proposing novel molecular structures and screening large virtual libraries, significantly expanding the number of candidate compounds that can be evaluated. Meanwhile, predictive models for absorption, distribution, metabolism, excretion, and toxicity (ADMET) are used to assess and prioritize candidates based on predicted drug-like properties before physical synthesis takes place.
Where researchers once faced the challenge of evaluating millions of potential targets and compounds through repeated rounds of physical testing, AI can now evaluate vast datasets and generate promising predictions in a fraction of the time. The result is a broader, faster, and more expansive early-stage discovery process that increases the number of viable avenues for experimental investigation.
Why biological validation remains essential
While AI outputs can be highly informative, they remain, in large part, hypothesis driven. Biological systems are complex in ways that current models do not fully represent. A compound that looks promising in silico may behave very differently in a cell, tissue, or whole organism. Off-target effects, unexpected toxicity, and context-dependent mechanisms of action are among the many variables that only reveal themselves through experimental work.
This is not a sign of AI being limited, but a reflection of the real-world complexity that biological systems introduce and the fact that AI outputs still require confirmation in biologically relevant contexts. Even the most sophisticated predictive models are trained on existing data, which means they are constrained by what has previously been observed. As a result, they tend to be better at recognizing known patterns than at anticipating genuinely novel biological behavior.
Regulatory frameworks further reinforce the role of experimental biology, requiring evidence at multiple stages of drug development. In vitro and in vivo data remain a core requirement for regulatory approval, rather than optional extras that could be replaced by computational methods alone.
Furthermore, some of the most important drug discoveries have come from unexpected observations in the wet lab, results that no computational model predicted and that only became visible through hands-on experimentation.
Lab-in-the-loop
What is emerging in practice is not a competition between AI and the wet lab, but a Lab-in-the-loop approach where computational and experimental workflows are tightly connected. AI narrows the search space and generates testable hypotheses, while the wet lab confirms, refines, and, where necessary, overturns them. Crucially, the data generated through biological validation then feeds back into computational models, improving their performance over time.
This distinction matters because the “replacement” narrative does not fully reflect what is happening. The industry is not moving toward a future where algorithms make experimental biology redundant. It is moving toward workflows where AI expands the scale of exploration while the wet lab expands the capacity to validate, characterize, and learn from what AI generates. Researchers who can work across both are becoming increasingly valuable in drug discovery programs.
According to the William Blair analysis,1 investment is flowing into platforms built around this iterative, closed-loop model, combining computational insight with experimental validation at each stage of the discovery process.
Future outlook
For research organizations, AI and wet lab capabilities need to be developed in parallel, not in sequence. Generative AI tools for lead discovery, for example, only create value if there is experimental capacity to validate what they produce. Equally, wet lab workflows that are not designed to feed data back into computational models are missing a valuable opportunity to improve future performance.
AI is not replacing the physical work of drug discovery, but it is reshaping the scale, speed, and nature of discovery outputs. And so, the more relevant question is not whether AI will replace the wet lab, but how organizations can build the infrastructure, skills, automation capabilities, and workflows needed to integrate computational and experimental approaches effectively.
Built on decades of drug discovery screening expertise and technology competencies, Revvity is a strategic partner to address unmet needs in drug discovery, supporting high-quality biological validation and helping bridge the gap between AI predictions and biological reality at scale through a Lab-in-the-loop approach that connects computational insight with experimental validation.
For research use only. Not for use in diagnostic procedures.
Reference
- From Code to Clinic: How AI is (and isn’t) rewriting the life of a drug | William Blair. (2026, June 2). https://www.williamblair.com/Insights/From-Code-to-Clinic-How-AI-Is-and-Isnt-Rewriting-the-Life-of-a-Drug
