Increasingly, the pharmaceutical industry is leveraging artificial intelligence (AI) to help accelerate data mining in drug discovery and research. AI-driven data is being used to gain insights from multiple data sets across research to discover new molecules, process data, and automate workflows more efficiently. AI is also being explored in drug development to address challenges and improve clinical trial processes and recruitment.
 
These past couple of years Big Pharma has embarked on numerous strategic AI alliances and investments aimed at boosting drug discovery efforts and creating efficiencies across R&D. For example, AstraZeneca’s partnership with BenevolentAI for drug discovery in systemic lupus erythematosus and heart failure, has allowed AstraZeneca to discover two additional novel AI-generated targets for chronic kidney disease and idiopathic pulmonary fibrosis. 
 
Additionally, last year Sanofi committed potentially as much as $5.2 billion under a research collaboration and license agreement with Exscientia to develop as many as 15 novel small molecule candidates across oncology and immunology, leveraging Exscientia’s AI-driven platform. 
 
Importantly, as more and more AI partnerships arise in the pharmaceutical industry, regulators are looking to establish safeguards around this powerful technology. 
 
Elligo Health Research, a healthcare-enabling research organization, recently introduced DataAI Connect, a new data and technology platform that aims to create efficiencies across the clinical study workflow leveraging data to accelerate clinical trials. By bringing together data, processes, and clinical expertise, the platform is designed to source, process, analyze, and distribute patient real-world data (RWD).
 
Elligo’s Chief Data Officer, Michael Ibara, Pharm.D., discusses the use of AI in clinical trials, potential roadblocks, and where we can expect to see advances. –KB 
 
Contract Pharma: In what areas are you seeing increased use of AI in clinical trials?
 
Michael Ibara: There are therapeutic-area specific applications of AI that are adopted in clinical trials, such as using machine learning to help classify radiographs or assess brain scans, but when we look at applying AI directly to the clinical trial workflow, one of the first areas we’re seeing real use of AI is in finding qualified patients for a given set of inclusion/exclusion criteria. There’s clear utility for using computers to search through a mass of electronic healthcare data to match clinical variables to a set of criteria — it saves time, looks through hundreds of pages of data, and can continue to improve based on past experience. 
 
Contract Pharma: Are you seeing any advances in trials as a result?
 
Michael Ibara: It’s quite early in the application of AI to see large-scale improvements, but we have already seen an advance in our ability to find patients at scale, and to be more accurate in determining if a patient is qualified for a study – using actual medical history vs relying on patients to recall or on investigators to extrapolate based on a few medical findings. AI excels at using a large and detailed set of data to match patients using a complex set of criteria. 
 
Contract Pharma: What areas remain a challenge incorporating AI?
 
Michael Ibara: The entire clinical trial workflow has areas that can be improved through the judicious application of AI in whatever form is best, e.g., using NLP, machine learning, or other techniques. Since the trial process is rife with regulatory and administrative tasks, these all are potential areas of improvement using AI. The key challenge here is not a technical one — rather it’s in educating the decision-makers about AI and the potential value it brings. A lack of technical understanding leads to both overly-optimistic and overly-pessimistic judgements, either of which will hinder real progress. 
 
Contract Pharma: How might those challenges be overcome?
 
Michael Ibara: Using AI to make real improvements in the clinical trial process often requires changing an outdated process that limits potential improvements, e.g., requiring a human review of a document when an AI approach could do it faster and more accurately. The fear of regulatory reprisals coupled with a shallow understanding of the technologies used in AI is a real roadblock to progress. This can only be addressed by a willingness of decision-makers to seek out and learn about solutions using AI that are grounded in solid clinical and regulatory knowledge, and that leverage AI to ease the burden on humans, execute more rapidly, and scale beyond human capabilities. If we know where to look, we’ll find that real progress in improving clinical trials through use of AI is being made today.