Perhaps the most widely publicized use for artificial intelligence (AI) and machine learning in life sciences is in drug discovery and development. New analytical approaches to the design and development of novel therapeutics and small molecules seem to be popping up every day, and investors are flocking to the value proposition. The cost of discovering, developing and bringing a new drug to market typically falls between $2 billion and $3 billion1, and remains one of the largest barriers to success in the industry. Reducing this cost with more targeted approaches to discovery and streamlining process workflows in research and development could significantly improve the profitability of these ventures.
The drug discovery process is massively data intensive, making it a natural application for the tools of the AI revolution. In the past, scientists have largely used manual processes in a trial-and-error fashion while pharmaceutical companies employ databases of millions of compounds and molecular designs. Sorting, filtering and gleaning information from these data libraries takes a significant amount of time, even with assistance from machine automation and robotics. AI offers the ability to sort and cross-reference these libraries for targeted results and faster discovery.
For example, a team in Germany developed a novel approach to discovery using a deep learning AI tool to query and cross-check millions of organic chemical reactions2. This led to the development of a detailed multi-step synthesis that would have taken human researchers much longer to construct. Other companies are using the technology to mine and query historical information and results to better predict clinical design pitfalls, as well as to facilitate more effective targeting of drugs for specific disease categories. For instance, Microsoft and Eagle Genomics are partnering to create an enterprise research platform to process large amounts of data on how bacteria, fungi and viruses within the body play a role in disease3.
In addition to direct discovery efforts, an entirely separate data challenge is present—one of process workflow, prioritization and pipeline management. Pharma companies typically work on many potential new drugs at once, and these efforts use multiple complex workflows, including sequencing and molecular engineering, validation, mapping, and inventory management integration. AI can help standardize and streamline data, and—most importantly—integrate that data with workflow management across these disparate processes. This improves speed and efficiency, which ultimately reduces drug pipeline management costs.
With this in mind, the use of AI in drug development does not come without challenges. One of the most glaring is the standardization of data, both within individual organizations and across companies nationwide. With so much data stemming from distinct processes, the development of common platforms and unified data protocols is needed before any tangible result can be recognized. But efforts are underway to better coordinate and unify data across companies. For example, Merck announced a collaboration with Accenture to utilize Amazon Web Services to develop a cloud-based platform that allows researchers to share data, which will speed up the process and reduce barriers to entry4.
Another obstacle for using AI is workforce education and talent management. Today’s medicinal chemists are often not familiar with the current information technology advancements and applications. Companies of all sizes need to invest in education and talent development, all the way from discovery to clinical development and beyond. They must encourage and nurture a multi-tooled workforce in the lab to successfully use these AI and machine learning technologies.
The promise of AI in drug discovery and development is immense, and the ultimate goal—more effective and affordable medicines—is on the horizon. Even with challenges to success, startups and large pharmaceutical companies alike have recognized the potential of these applications, and significant efforts are underway to deploy them in the field. It is an exciting time in the field of drug development, with new advancements and initiatives being announced every day, and AI will undoubtedly play a role in developing new medicines for us all.
In part two of our series on the role of AI and machine learning in life sciences, we will explore how these applications can help in another step in the development process: clinical trial management.
1“Cost of Clinical Trials for New Drug FDA Approval Are Fraction of Total Tab,” John Hopkins Bloomberg School of Public Health, Sep. 24, 2018. 2“Need to make a molecule? Ask this AI for instructions,” Nature, March 28, 2018. 3“Microsoft turns to the microbiome with Eagle Genomics partnership,” Fierce Biotech, Oct. 15, 2018. 4“Amazon, Merck, Accenture to launch data-driven drug development platform,” Becker’s Hospital Review, Sept. 19, 2018.