Artificial intelligence in drug discovery: what is realistic, what are illusions? Part 1: Ways to make an impact, and why we are not there yet

Drug Discovery Today. 2021, 26, 2, 511-524

DrugDiscovery AI realistic paper 2021

Artificial intelligence (AI) is having a moment in the limelight, but what does this mean in the field of drug discovery? In this review, Bender and Cortés-Ciriano cut through the hype surrounding AI and provide a high-level assessment of its current application in drug discovery. Crucially, the review highlights the necessity of making better-quality decisions versus improving the speed and cost at each phase of the drug development process, measured in terms of project success (net-profit). With this in mind, the authors suggest that efforts to develop AI should focus on the reduction of failures through improved decision-making, and discuss the challenges in doing so with the currently-available data and culture within the pharmaceutical industry. While the current approach of isolating a disease-driving mechanism and target, followed by the collection of proxy data (e.g ligand activity against a protein), provides large amounts of data for AI model training, it does not necessarily answer the important questions that lead to high quality drug candidates, such as probing the overall efficacy and safety of the compounds. These questions are much more difficult to address due to the complexity and multi-dimensionality of in vivo systems. A significant barrier to fully utilising the wealth of AI methods that are currently being developed, therefore, is the collection and curation of high-quality, biologically-relevant data. The authors also comment on the difficulty in coordinating data-sharing consortia, considering the need for confidentiality and competitive edge within the pharmaceutical industry, as well as the inconsistencies present between data collected from different laboratories.

Overall, this review provides an insightful read that prompts the community to rethink the way we collect and utilise data for drug discovery. We look forward to Part 2, which will discuss the relevance of chemical and biological data for the application to AI in more detail.