The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Machine Learning at GSK with Kim Branson - #536
Nov 15, 2021
Kim Branson, SVP and global head of AI and machine learning at GSK, shares insights on leveraging machine learning in pharmaceuticals. He discusses the integration of genetics data for drug discovery and the impressive 500 billion node knowledge graph designed to mine scientific literature. Kim also highlights their recent collaboration with King’s College, focusing on personalized cancer research using AI. The conversation dives into the challenges of building scalable AI infrastructures and the critical need for robust evaluation programs in real-world applications.
AI Snips
Chapters
Transcript
Episode notes
GSK's Three-Pronged Strategy
- GSK leverages genetic data, functional genomics, and AI/ML for drug discovery.
- Genetics provides clues, functional genomics validates targets, and AI/ML integrates the data.
Genetics and Functional Genomics
- Genetic databases offer clues about potential drug targets, but further experimentation is needed.
- Functional genomics helps determine a target's validity by observing the effects of gene perturbation.
Variant-to-Gene Mapping Model
- Kim Branson discusses a variant-to-gene mapping model using a ranking approach, incorporating diverse features.
- This model iteratively improves by incorporating experimental feedback, increasing the percentage of mappable variants.