The average number of years it takes to bring a drug candidate from phase I clinical trials to regulatory approval is around 10.5. When you employ an effective biomarker strategy, this time can be cut down to approximately 9.1 years, and perhaps even further using the power of AI. In this presentation, Yiannis Kiachopoulos, CEO and Co-founder of Causaly, introduces how AI can assist scientists in biomarker discovery, addressing key problems in the process. Every day counts for patients waiting for new drugs.
First, Kiachopoulos characterised biomarker discovery as a ‘knowledge problem’. There are many different types of biomarkers and a vast amount of scientific literature to process. Furthermore, biomarker discovery suffers from what computer scientists call a ‘long tail problem’. The information that bioinformaticians are looking for usually lies in the scarcer, more niche publications, leading to fewer biomarkers translating into the clinic. This data also suffers from bias when scientists need to progress or drop a candidate. Finally, there’s poor traceability in the development process due to insufficient collective memory within an R&D organisation.
Considering how AI can be used to overcome these problems, Kiachopoulos stressed that AI should principally be used as a tool to help scientists, rather than the other way around. AI should support human scientists by processing vast amounts of data and providing relevant information without introducing bias.
Kiachopoulos provided examples of how AI can help identify and validate biomarkers, streamlining the research process for scientists. He then presented a case study from a colleague demonstrating how AI can be used to identify diagnostic biomarkers for luminal A breast cancer. The case study showed that AI can generate hypotheses, analyse pathways, and provide detailed information on biomarker expression, significantly aiding research.
The presentation concluded that AI can greatly enhance the efficiency and effectiveness of biomarker discovery, but it should always support and not replace human scientists.
