Data Analysis and Informatics

Preparing Early Drug Discovery for the Machine Learning Revolution

• SH

  Scott Harrison, PhD

  Principal Scientist
  Merck

The mass-collection of data and its mining with machine learning algorithms is driving a revolution within the industrialized world, enabling data-forward organizations to improve operational efficiency and facilitate decision-making. Operationalizing data science and machine learning pipelines requires well-behaved, curated datasets on which to train models, and ideally, a streamlined workflow that handles data collection, standardization, model building and publication. In early drug discovery, the process of iterative drug design known as the “design, make, test” cycle, whereby chemists design and evaluate drug candidates, synthesize selected designs, and test them in biological assays is becoming increasingly complex as the technologies employed in each step continue to advance. Successful management of these complex workflows is often difficult to accomplish with available commercial software packages, which often necessitates the construction of novel applications that are capable of integrating with legacy systems. While building highly-integrated software tools is a distinct challenge, the reward is an opportunity to gather information which could be used to further improve the efficiency of drug discovery. This talk will highlight Merck’s approach to building tools that enable collaborative drug design and high-throughput chemical synthesis while gathering information that we anticipate will transform the way scientists discover new medicines.