Healthcare innovation in drug development intersects with a few critical trends: the rise of artificial intelligence (AI) applications in healthcare—particularly machine learning (ML) and predictive analytics; the need to reduce the costs associated with drug development; and the need to improve the performance of these treatments. A discussion paper from the U.S. Food and Drug Administration published in May seeks comment from industry stakeholders on the applications of AI/ML in drug development, also referred to as pharmatech. The paper lays out a compelling overview of how digital tools support drug development, particularly because they are capable of aggregating different types of data.
Check out conversations around pharmatech innovations and other healthcare topics at HLTH 2023 in Las Vegas October 8-12. To register, click here!
The paper, “Artificial Intelligence and Machine Learning in the Development of Drug and Biological Products,” highlights how pharma and biotech companies, regulators, academic groups, and other stakeholders are working to develop a shared understanding of how specific innovations in AI and ML can be applied throughout the drug development process. The comment period is scheduled to end August 9.
Here is a look at just a few of the ways AI/ML is being applied to drug development.
Identify drug development targets
For example, AI/ML may be applied to pools of aggregated scientific data— such as genomic, transcriptomic, proteomic, and other data sources from healthy people as well as those with a specific disease— to provide a significant opportunity to inform biological target selection for drug development, according to the report. Biopharma companies have used AI for 24 drugs currently in or poised to enter clinical trials and that number continues to rise, according to a story published by MIT Technology Review.
Recursion Pharmaceuticals and Causaly represent two companies identifying biomarkers for drug development that received hefty backing from venture capital and strategic investors earlier this month. Causaly, a London-based life science tech company that counts 12 pharma companies as customers, raised $50 million in a Series B round led by ICONIQ, with participation from Index Ventures, Marathon Venture Capital, EBRD, Pentech Ventures and Visionaries Club.
Causaly’s work involves using a knowledge graph with advances in generative AI to help researchers conduct deep, unbiased scientific exploration, according to a company press release. The technology is being adopted at scale by teams of researchers in diverse workflows from target identification to biomarker discovery. Customers include Gilead, Novo Nordisk, and Regeneron, among others.
Recursion received $50 million as part of a collaboration deal with Nvidia, which will provide its cloud platform for Recursion to train its AI models. In May, Recursion acquired two companies to accelerate its AI-fueled drug development business — Cyclica and Valence. Recursion currently has drug research partnerships with Roche and Bayer.
Insilico Medicine at the end of June became the first pharma company with a drug generated by AI to enter Phase 2 clinical trials, according to the company. The drug targets idiopathic pulmonary fibrosis, a rare disease. Insilico analyzes data to discover disease signatures and to find promising targets for already existing molecules or new ones that can be designed. It has three goals, according to an article from MedCity News senior biopharma reporter Frank Vinluan: Speed up disease target identification, generate novel information on molecules, and predict clinical trial outcomes.
Clinical trial recruitment
Clinical trial recruitment is another area where life science startups apply AI. Massive Bio has collaborated with CureMatch to combine their expertise in personalizing cancer treatments with analyzing medical data to match patients with clinical trials. At the same time, they are also using their collaboration to improve diversity in clinical trials to ensure that cancer therapeutics maximize effectiveness across diverse patient populations.
The FDA has witnessed a rapid growth in submissions that reference AI/ML across drug and biological product applications. In 2021 these submissions rose to 100. They reflect a diverse range of therapeutic areas. The applications not only include drug discovery and clinical trial enrichment but also endpoint assessment and post-market safety surveillance, according to the FDA discussion paper.
Modeling drug impact on humans and animals with organ on a chip
Although organ-on-a-chip technology has not always had an AI component, it is intended to model the impact of drugs on various parts of the body, so predictive analytics are not a stretch for this platform technology. Emulate has developed several different organs on a chip including one to assess a drug’s impact on the brain, kidney, liver, lung, and parts of the intestine, such as the duodenum and the colon. For drugs that have a high rate of failure—particularly oncology drugs, of which only 5% of drugs that make it to Phase 1 succeed—organ on a chip tech provides an affordable way to assess complex therapeutics on the body.
Predicting side effects
AI modeling can also be used to predict drug side effects. Data scientists at the Icahn School of Medicine at Mount Sinai in New York created an AI model to predict which drugs have the potential to produce congenital disabilities, even though they are not currently identified as harmful.
There is also encouraging research on the potential for AI to identify and prevent adverse drug events when certain drugs are combined, according to a study published in The Lancet. Adverse drug events can be costly and impact patient health in the short- and long-term. Unfortunately, few studies assess AI in clinical settings and most studies were published in the past 5 years. We should expect robust studies of AI in assessing and preventing adverse drug events as an area of focus in the years to come.
Although digital platforms for AI-enabled drug development are a compelling area of pharmatech, approaches to this tech remain in the early stages of development and will require more clinical research. One of the challenges of aggregating data is that the datasets are usually complex and come from disparate sources, which can make it tricky to standardize. But the potential to transform drug development, shorten the time it takes from bench to bedside and improve the effectiveness of drugs for a wider patient population will be the long-term gains of this tech.
Check out conversations around pharmatech innovations and other healthcare topics at HLTH 2023 in Las Vegas October 8-12. To register, click here!
Photo: Yuuji, Getty Images
