By Matthew Chun
In recent months, generative artificial intelligence (AI) has taken the world by storm. AI systems like ChatGPT and Stable Diffusion have captured the imagination of the masses with their impressive and sometimes controversial ability to generate human-like text and artwork. However, it may come as a surprise to some that — in addition to writing Twitter threads and dating app messages — AI is also well underway in revolutionizing the discovery of life-saving drugs.
Milestones in AI-Enabled Drug Discovery
Far from being a distant sci-fi future, AI-enabled drug discovery is already here. A non-exhaustive list of historic milestones in the field includes the following achievements:
- In early 2020, Exscientia announced the first-ever AI-designed drug molecule to enter human clinical trials.
- In July 2021, an AI system by DeepMind called AlphaFold predicted the protein structures for 330,000 proteins, including all 20,000 proteins in the human genome. The AlphaFold Protein Structure Database has since expanded to include over 200 million proteins, covering nearly all cataloged proteins known to science.
- In February 2022, Insilico Medicine reported the start of Phase I clinical trials for the first-ever AI-discovered molecule based on an AI-discovered novel target—all done at a fraction of the time and cost of traditional preclinical programs.
- In January 2023, AbSci became the first entity “to create and validate de novo antibodies in silico” using generative AI.
- In February 2023, the FDA granted its first Orphan Drug Designation to a drug discovered and designed using AI; Insilico Medicine plans to begin a global Phase II trial for the drug “early” this year.
According to Boston Consulting Group, as of March 2022, “biotech companies using an AI-first approach [had] more than 150 small-molecule drugs in discovery and more than 15 already in clinical trials.” But how exactly is AI being used to accomplish these milestones, and why does it matter?
How AI Is Being Used
Traditional drug discovery is a notoriously time consuming and expensive process, with pre-clinical stages typically taking three to six years and costing hundreds of millions to billions of dollars. However, a host of AI tools are revolutionizing nearly every stage of the drug discovery process, offering substantial potential to reshape the speed and economics of the industry.
- Target identification: At the target identification phase of drug discovery, AI is being trained on large datasets, including omics datasets, phenotypic and expression data, disease associations, patents, publications, clinical trials, research grants, and more to understand the biological mechanisms of diseases and to identify novel proteins and/or genes that can be targeted to counteract those diseases. Combined with systems like AlphaFold, AI can go even further than mere target identification by predicting the 3D structures of targets and accelerating the design of appropriate drugs that bind to them.
- Molecular simulations: AI is also being used to reduce the need for physical testing of candidate drug compounds by enabling high-fidelity molecular simulations that can be run entirely on computers (i.e., in silico) without incurring the prohibitive costs of traditional chemistry methods.
- Prediction of drug properties: Some AI systems are being used to bypass simulated testing of drug candidates by predicting key properties such as toxicity, bioactivity, and the physicochemical characteristics of molecules.
- De novo drug design: While traditional drug discovery has historically involved the screening of large libraries of candidate molecules, AI is shifting this paradigm too. Some systems are capable of generating promising and never-before-seen drug molecules entirely from scratch.
- Candidate drug prioritization: Once a set of promising “lead” drug compounds has been identified, AI is used to rank these molecules and prioritize them for further assessment, with AI approaches outperforming previous ranking techniques.
- Synthesis pathway generation: Going beyond theoretical drug design, AI is also being used to generate synthesis pathways for producing hypothetical drug compounds, in some cases suggesting modifications to compounds to make them easier to manufacture.
As AI systems continue to improve, the idea of fully automated end-to-end drug discovery appears less and less to be matter of if, but of when.
A Growing Industry
The excitement for AI-enabled drug discovery extends beyond just scientists, with investors taking notice as well. According to Morgan Stanley, even “modest improvements in early-stage drug development success rates enabled by the use of artificial intelligence and machine learning” could result in an additional 50 novel therapies over a 10-year period, representing a more than $50 billion opportunity. Others appear to agree, with third-party investment in AI-enabled drug discovery more than doubling annually for five consecutive years and reaching more than $5.2 billion at the end of 2021. A selection of recent financings from February 2020 to April 2021 reveals a number of players, including Schrödinger, Insitro, AbCellera, Relay Therapeutics, Atomwise, Recursion Pharmaceuticals, XtalPi, and ExScientia, who have all raised hundreds of millions of dollars to pursue their AI-driven drug discovery pipelines.
If current trends continue, it will only be a matter of time before the drugs we take are no longer designed by people, but by machines. With the promise of lower costs and shorter development timelines, AI-enabled drug discovery holds massive potential to increase the accessibility of drugs and to treat presently incurable conditions. However, it also opens the floodgates to a host of unresolved issues relating to, e.g., intellectual property rights, the risk of technological misuse, and the continued assurance of drug safety and efficacy in this new era.
Will we be ready to seize the opportunity, or will we get mired in the challenges? Our preparation as lawyers and policymakers must start now, because the future of AI-enabled drug discovery is already here.