AI for drug discovery is no longer a distant dream — it’s a reality. Here are the 11 AI companies in Phase 1/2 clinical trials…
History of integrating technology in drug discovery as a field:
The field of drug discovery has long been characterized by its significant costs and time-consuming nature. The small biotechs who take on this challenge invest 10–15 years and up to $1–2 billion dollars on projects with a >90% failure rate. However, the integration of technology, particularly AI, into the drug discovery process has the potential to usher in a new era of efficiency and innovation.
While the concept of incorporating technology into drug discovery is not new, previous attempts did not live up to their initial hype. For example, the Watson supercomputer, famous for its Jeopardy! victory in 2011, was positioned as a revolutionary tool for healthcare and drug discovery. However, early efforts to utilize Watson in the medical field were underwhelming as the supercomputer gave advice that was incompatible with patients’ medical histories, recommended treatments that did not exist in parts of the world, and struggled to modulate the inherent messiness of clinical data. In 2022, Watson was sold-off for parts at a meaningful discount.
The rapid advancements in AI technologies since (including a step-function improvement in large language models in 2023), coupled with the exponential growth of available biomedical data, have paved the way for its practical application in the pharmaceutical industry. Scientists and researchers are now able to navigate the vast ocean of biomedical data more efficiently than ever before.
Leading use cases for AI in drug discovery and development:
AI has a wide range of use cases across the entire drug development timeline, encompassing various stages from research and development to clinical trials, manufacturing, and commercialization. For purposes of this article, the focus will be on the discovery and development integrations (#1–3).
- Drug target identification and validation: AI can help analyze large genomic datasets, generate novel knowledge graphs, map genomic pathways and networks, and assist with prioritizing and validating of drug targets
- Drug design and optimization: AI can be used to optimize drug design by analyzing the binding affinity, selectivity, and pharmacokinetic/ pharmacodynamic properties of promising molecules
- Virtual screening and compound generation: AI can be employed to virtually screen vast libraries of chemical compounds against specific targets and identify drug candidates with desired properties; similar screens can also be used to identify off-target binding and/or toxicity concerns
- Generative chemistry and synthesis planning: AI can help generate novel chemical structures with desirable properties and assist in optimizing the synthesis planning process, enhancing efficiency in the production of target compounds
- Drug repurposing and repositioning: AI can be used to identify new therapeutic applications for existing drugs that have already undergone clinical testing, including evaluating interventions at different steps of the genomic pathway or treatment in adjacent therapeutic areas
- Clinical trial optimization and patient stratification: AI could be used to improve clinical trial design, patient recruitment, and trial optimization. AI models can also analyze patient data to identify subgroups that are more likely to respond to treatment
- Data integration and analysis: AI can help researchers analyze, interpret, and extract insights from clinical and commercial datasets
Who are the leading AI for drug discovery players:
The field of AI for drug discovery has seen several frontrunners emerge. Recursion, Insilico, and Relay all have AI-discovered drugs in Phase 2 development, while others including Insitro, Isomorphic Labs, and Valo Health have raised large financing rounds on their AI platform technology. Common platforms for these companies have emerged:
- Focus on genetic disease: The genetic basis provides a clearer understanding of the underlying molecular mechanisms and enables companies to leverage large-scale genetic datasets to accelerate the discovery and development of novel therapies. (ie. NF2, CCM, FGFR)
- Therapeutic area agnostic: TA-agnostic approach in AI drug discovery allows companies to capitalize on the flexibility, scalability, and broad applicability of their technology. They strive to first identify potent molecules along select genetic pathways and then optimize for disease-specific dynamics (ie. transcriptionally addicted cancers, opioid dependence)
However, AI is not a magic bullet. In May 2023, Benevolent AI’s lead asset failed a Phase 2a trial in atopic dermatitis and shares tumbled 79%. Earlier in April, Relay’s PI3K inhibitor was successful mechanistically but failed to translate into clinical benefit, sending the stock down 36%. What’s more, the majority of these AI drug discovery companies are still pre-clinical or in early discovery (ie. target identification, lead optimization).
“The role of AI in drug development is not to guarantee success, but to help us fail faster and fail less frequently.”
AI drug discovery companies will inevitably face the same challenges as traditional large pharma. Early setbacks have been faced by all new modalities within biotech (ie. gene therapy, CRISPR/Cas9) before scientists learned how to properly utilize the new technology. Given enough time, the micro-optimizations enabled by AI across all stages of drug development should translate into meaningful clinical benefit. This could dramatically reshape the drug development cycle to be faster, safer, and more efficacious, resulting in substantial cost savings that in turn fund new research and development. AI has the potential to bring new drugs to market that may otherwise never have been discovered.
