Published Jan 10, 2023

643: A.I. for Medicine — with Charlotte Deane

Jon Krohn and Charlotte Deane delve into the transformative role of AI in medicine, exploring its application from antibody diversity and evolution to innovative approaches in epidemiological modeling, personalized medicine, and accelerated drug discovery in biologics.
Episode Highlights
Super Data Science: ML & AI Podcast with Jon Krohn logo

Popular Clips

Episode Highlights

  • Biologics

    , Chief Scientist of Biologics AI at Exscientia, explains the role of biologics in modern medicine. Unlike traditional small molecule drugs, biologics are protein-based therapies, such as antibodies, which are used to treat serious conditions like cancer and COVID-19 1. These treatments are often administered via injections or IV drips due to their complex nature, which makes them unsuitable for oral consumption 2. Despite their effectiveness, biologics are expensive and slow to develop, prompting interest in computational methods to accelerate their discovery and reduce costs 3.

       

    Antibody Design

    Designing antibodies using AI involves optimizing their binding to specific targets while maintaining therapeutic properties. highlights the complexity of this task, noting the vast diversity of antibodies naturally present in the human body 4. The goal is to create antibodies that bind effectively to disease-causing proteins without triggering adverse immune responses 5. This requires a multi-faceted approach, balancing factors like manufacturability and stability, to ensure the antibodies can be safely stored and administered 6.

       

    Protein Prediction

    AI plays a crucial role in predicting protein structures, a task that is particularly challenging for antibodies due to their complex binding dynamics. and discuss the advancements in this field, including specialized models like AlphaFold that improve prediction accuracy 7. The emerging challenge is understanding 4D protein dynamics, which involves how protein structures change over time during interactions 8. Adjusting loss functions in AI models is essential to focus on the most variable parts of the protein structure, enhancing the precision of these predictions 9.

Related Episodes