Abstract

The pharmaceutical industry has had a long-standing reputation of being extremely data-rich and algorithm poor, which has resulted in a lag in exploiting advances in AI when compared to other major industries. Over the last five years machine learning-based approaches have started to receive more attention, with investments reaching $13B in 2020, a four-fold increase over 2019. This type of steady and increasing investment has taken the space from a small number of sub-critical mass players to a healthy ecosystem of large and small players to complement the activities in the established pharma companies. There are countless examples of successes from target ID through drug development where machine learning models have made significant contributions. These managed to disarm many skeptics, but more importantly generated tangible pipeline value. However, their impact was typically confined to the traditional stagegates of drug discovery and development. For the first time, the scope, scale and quality of human data coupled with computational power can enable a foundational shift in the approach to developing new medicines throughout the entirety of drug discovery and development. The next phase of our industry’s journey to become more data-driven will be to use AI and data to break down the lines between the traditional chevron model of pharma. This human-centricity will naturally create a new series of challenges around integration of diverse skillsets, around data biases, and around data sharing. Understanding and addressing these challenges will impact not just the value of the work within each chevron, but finally enable us to develop models to meaningfully break the silos and impact downstream attrition and ultimately the number and quality of drugs reaching patients.