Even though AI-designed drugs aren’t yet a household term for FDA-approved, commercially available therapies, they are a reality in clinical development pipelines. In spring 2024, CBER reported more than 70 investigational new drug applications (IND) that involved artificial intelligence (AI) or machine learning (ML) in some way.
Much of that progress is because of where AI is currently winning in drug discovery. While there are several “what-if” opportunities for future preclinical innovation, there are also many use cases where AI is successfully applied at the bench right now—at multiple junctures of drug discovery workflows.
“We have seen remarkable progress as far as AI’s impact on drug discovery, from target identification and drug design to preclinical safety and beyond—and this innovation has catalyzed and will continue to catalyze new therapeutic possibilities,” said Rory Kelleher, Global Head of Business Development, Healthcare and Life Sciences at NVIDIA. “In the future, I think what you’ll see is not necessarily a distinction between human-designed drugs or AI-designed drugs, but simply an era where intelligent technologies play an inherent, complementary role in drug discovery.”
Owing in great part to the speed and capacity that machines have compared with humans, AI and ML have already helped accelerate preclinical research—while at the same time addressing many of its operational setbacks such as costs and labor availability. Some experts estimate that with AI’s support across scientific excellence and efficiency, sponsors might bring to market 50 more therapies than they otherwise would have over 10 years.
“What’s especially rewarding about working in this field is that we’re seeing the scientific benefits materialize right alongside operational ones,” Anthony Costa, PhD, Director of Digital Biology at NVIDIA said. “When you look at the speed that advanced computing is enabling across data analysis, experiment execution, trial simulation, and other activities in drug discovery, those dual strengths are absolutely helping innovate preclinical work at the same time they’re operationalizing it. And the world, inevitably, will benefit from that as these assets make their way to the commercial market.”
Here are four use cases that best exemplify those already-established impacts of AI on drug discovery:
1. Target identification and mechanism-of-action (MOA) delivery[1] [2]
Applications of AI that support target identification and MOA are increasingly becoming part of the drug discovery ecosystem. In some cases, these applications identify novel targets—like a TNIK inhibitor which was identified through AI, for example.
Other cases involve repurposing existing products into newly approved indications. For example, a knowledge graph was used in early 2020 to search for relevant MOAs for COVID-19. During that work, researchers identified the rheumatoid arthritis drug baricitinib as the strongest candidate from thousands of others. Originally granting an emergency approval for COVID-19, the FDA converted the drug’s status into a full approval in 2022.
2. Hit identification
Neural network training has helped to expedite the speed at which hits are identified, having already supported drug discovery for antibiotic resistance, as just one notable example. The antibiotic halicin (originally a diabetes candidate) was found that way, after being evaluated against multiple bacterial strains to identify assets that could be repurposed.
3. Drug design and optimization
The expanding market for commercial platforms such as NVIDIA’s BioNeMo and others has become a resource for biotechs across small molecule design, protein design, RNA/genetic medicine development, and other uses to help sponsors prioritize, optimize, and design their candidates. The ranking of identified molecules has outperformed traditional approaches, for example.
4. Preclinical safety
AI has also been employed to predict ADMET profiles, absorption, distribution, metabolism, excretion, and toxicity. For example, one platform—aptly named ADMET-AI—does just that. Available to the public, the platform was trained on dozens of ADMET datasets and features a website that allows input of up to 1,000 molecules to make predictions.
Setting the foundation for ongoing innovation
As these and other use cases continue to demonstrate what AI can do at all corners of drug discovery, a prevailing win is not just about how the tech is advancing life sciences knowledge—but also how it can make those advancements more sustainable long-term despite market challenges and dynamics.
In a biotech industry battered with funding volatility, talent gaps, and pressures from a 94% drug discovery failure rate, being able to do more with less—while also getting preclinical workflows done faster—will be key. Given AI’s proven ability to rapidly analyze data, identify trends, and unlock predictive intelligence, these “current wins” are just the start, Kyle Tretina, PhD, Digital Biology Product Lead at NVIDIA says.
“These use cases are just scratching the surface of what’s possible when AI is integrated into life sciences. They’re setting the stage for a future where breakthroughs aren’t constrained by time or cost in the same way. AI isn’t just a tool—it’s a catalyst for reimagining preclinical R&D, enabling us to tackle complexity and scale in ways that were unimaginable just a few years ago”, says Kimberly Powell, Vice President of Healthcare at NVIDIA, “The advances we’re seeing today—such as faster drug discovery cycles, more precise modeling of molecular interactions, and smarter optimization of therapeutic candidates—are foundational. I believe we’re entering an era where the life sciences industry will fundamentally transform, driving sustainable, long-term advancements that improve lives globally.”