Functional Pearl: Signature Memoization for Drug Repurposing

on YouTube ↗28m

" summary

Claims that signature memoization caches subexpressions to cut edit/run/check cycles; implements a tiny embedded DSL Dagopoly in Scheme (233 lines) and Python (≈300 lines) that writes cached artifacts (pickled streams, human-readable manifests). Applies to Alzheimer’s drug repurposing by joining a 76-gene AD list to the RTX-KG2 graph, using reservoir/random sampling (2%) to get interactive results in ~45+30s for ranking and explanations versus ~12 minutes cold, reading a 4.5 GB tgz.

AI-written summary · about

" chapters (12)

1. 0:00Introduction
2. 1:50Drug repurposing concepts
3. 5:03FDA approval and Alzheimer's
4. 7:34Memory efficiency design
5. 9:01Generators streams block objects
6. 11:32Caching behavior API
7. 12:57Comparison with SnakeMake
8. 14:25Multilingual Python Scheme
9. 15:07Drug target metrics
10. 15:50Live demo
11. 21:15Q&A
12. 25:26Implementation summary