Octant and Schrödinger profile 700+ molecules to explain why pKa and logP predictions clash with experiments
_onionesque · x · 2026-11-01
A joint blog post by Octant Bio and Schrödinger analyzes 700+ molecules where experimental pKa/logP measurements disagreed with computational predictions.
- Disagreements often traced to tautomerism, reactive groups, or assay blind spots rather than insufficient data
- Authors argue understanding uncertainty sources matters more than simply collecting more data
- The accompanying dataset is openly released (DOI: 10.5281/zenodo.23345225) for community modeling
A substantive hands-on reference for molecular design, ADME prediction, and AI4Science modeling.
More from AGI Musings
- Stanford's Christopher Potts publishes long rebuttal to skeptical views of interpretability research — ChrisGPotts · 2026-11-01
- Vitalik Buterin pens 'Digital Minimalism' essay urging intentional, deliberate tech use — SuhailKakar · 2026-11-01
- NYT: Data scientist argues predictive algorithms are flattening culture into monotony — nordicinst · 2026-11-01
- Burkov: humans validating machine work is no different from using a tractor — burkov · 2026-11-01
- Lightcone operator publicly calls Scott Alexander's Anthropic strategy 'unethical' — ohabryka · 2026-11-01
- OpenAI staffer predicts full recursive self-improvement within 18 months — willdepue · 2026-11-01