Artificial Intelligence and Drug Innovation: A Large Scale Examination of the Pharmaceutical Industry

Bowen Lou, University of Connecticut, and Lynn Wu, Operations, Information, and Decisions, The Wharton School

Abstract: We study how artificial intelligence (AI) affects the drug development process. Despite substantial investment, the chemical novelty and efficacy of new drugs have declined. As AI becomes an important general-purpose technology, it has been touted for an ability to accelerate drug discovery. To understand how AI affects drug discovery, we not only examine AI’s effect on final outcomes but also distinguish its effects on different stages. We approximate firms’ AI capabilities by analyzing patents and job postings and find that AI can primarily affect the earliest stage in drug discovery when tasks are heavily dependent on automatic data processing and reasoning. However, AI is less useful later in the process when human engagements and judgements are required. We further find that despite finding more new compounds in the earliest stage, AI has not changed the total number of drugs achieving final approval by the FDA, suggesting most of these early compounds have not proven effective in later stages. We also find that AI is most effective in producing drugs at the medium level of chemical novelty but is less helpful with drugs at the extreme ends of the spectrum — those that are either entirely novel or incremental “me-too” drugs. Overall, our study sheds light on both the advantages and the limitations of using AI in drug development.

Read the full working paper here.