Protein folding and what it means for drug discovery
summary
Predicts that AlphaFold models approach experimental accuracy (AlphaFold 2 GDT >90), and that AlphaFold 3 adopts diffusion models; runs sequences through AlphaFold on Colab because local runs need terabytes and an NVIDIA GPU. Fetches structures from the Protein Data Bank, uses biopython, RDKit, NGLview/PyMOL for visualization, and docks ligands with Dock‑string + AutoDock Vina, producing scores like -6.1 (noradrenaline), -8.0 (propranolol) and -8.5 (a statin that docks energetically well but not in the correct pocket).
AI-written summary · about
chapters (14)
- 0:00Protein folding basics
- 1:07How proteins fold
- 3:16Why proteins as drug targets
- 5:39Proteins in cell signaling
- 9:54Beta1 adrenergic receptor
- 10:37Python tools for protein visualization
- 13:09Gene to protein with AlphaFold
- 16:25Protein databases exploration
- 18:08Drug design for binding sites
- 21:40Drug binding prediction
- 24:05Drug interaction validation
- 26:11Limitations of folding models
- 27:13AlphaFold's drug discovery impact
- 28:57Q&A
related talks
36mAlgorithmic Alchemy: Transforming Drug Discovery with Open AI
01h 13mDeep Learning in Structural Biology and Protein Design: How, Where, and Why
04mHow Machine Learning tools at Google made AlphaFold possible
53mHow machine learning is influencing protein engineering
10mGenesis Final Round: Nanyang Biologics
16mBeyond Patterns: The Case for Physics-Grounded AI in Drug Discovery
06mTools vs Products: Lessons learned from a niche industry
30mLeveraging Molecular Data for Drug Discovery