Machine Learning: A New Approach to Drug Discovery
summary
Asserts that a scaled bio-data factory combining induced pluripotent stem cells, CRISPR perturbations, high‑content imaging and lab automation produces datasets that let ML predict downstream human outcomes well enough to change go/no‑go decisions. Provides case studies: 16p11.2 CNV neurons show reproducible synaptic arborization shifts detectable by ML; a NASH cell assay separates patients and highlights lipid droplets; claims human genetics doubles approval odds (~2x) while R&D cost per approved drug exceeds $2.5B.
AI-written summary · about
chapters (18)
- 0:00Keynote Address
- 2:32Low R&D Success & High Costs
- 5:33More data enables rich models to achieve higher performance
- 6:26End-to-end models to uncover features
- 7:26Machine learning to make sense of high-content data
- 10:15Big data in biology: Convergence of revolutions
- 12:25Predictive Models for Drug Discovery
- 13:24Mice as a model for humans
- 14:13Humans as a model for humans
- 15:13Growing amount of phenotypic data
- 16:13Genetics suggests causality
- 16:27Growing number of genetic associations
- 17:16Genetic evidence increases approval odds
- 18:17Disentangling mechanism using cell biology
- 22:38Classification of Genetic Interventions
- 24:04A Cellular Model for NASH
- 27:38Epochs of science
- 29:13The next epoch
related talks
30mAI in Drug Discovery
30mLeveraging Molecular Data for Drug Discovery
20mDiagnostic Medical application of ML
35mFrom cells to drug responses - machine learning in cancer research
29mMachine Learning Techniques for Omics Data Analysis
53mThe Algorithm for Precision Medicine
08mEngineering the Future of Pharma: AI-Driven Drug Discovery
36mFormatting Biological Big Data to Enable (Personalized) Systems Pharmacology