Machine Learning-Based Design Of Proteins
summary
Machine learning trains a predictive model on assay data for a seven‑amino‑acid insertion in the AAV capsid, then optimizes a distribution over sequences (parameterizing marginal amino‑acid distributions per position) with an entropy constraint to trade off diversity versus expected viability. The optimized library halves the non‑packaging waste of standard libraries and produces downstream gains — a held‑out infection assay shows roughly 10× improved brain infectivity compared with the prior state of the art.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:18Protein engineering
- 3:17The combinatorial space
- 4:34Directed evolution
- 7:31Work synergistically
- 8:19Predictive models
- 11:38The problem
- 15:52Epistemic uncertainty
- 18:58Library design
- 19:27Real life example
- 20:33Optimization problem
- 23:21Algorithm description
- 24:30Language of probability
- 25:55Gene therapy
- 29:46How we did this
related talks
46mMachine Learning-based Design of Proteins and Small Molecules
53mHow machine learning is influencing protein engineering
30mMachine Learning: A New Approach to Drug Discovery
43mMachine Learning in Molecular Biology
31mML methods to accelerate the drug discovery cycle
01h 13mDeep Learning in Structural Biology and Protein Design: How, Where, and Why
16mBeyond Patterns: The Case for Physics-Grounded AI in Drug Discovery
28mMachine Learning vs The Flu