Nat. Comput. Sci. | Buchwald–Hartwig反应的稳健分布外预测
Ranking
Overall
79
Content
90
Popularity
55
Observed public metrics from 1 member.
Merged summary
TL;DR - A Nature Computational Science study shows that standardized, chemically diverse data plus active learning enables robust out-of-distribution prediction of Buchwald–Hartwig reactions. The approach improves reaction-condition selection for unseen substrates without requiring complex models.
- Researchers combined 11,300 new automated experiments with public high-throughput data, creating a standardized dataset of about 27,500 reactions.
- Their diversity metric, CRDS, correlated more strongly with strict OOD performance than dataset size did (Pearson 0.79 versus 0.60).
- A random forest exceeded 90% ROC AUC on high-confidence OOD predictions, highlighting the importance of data quality and representation.
- Testing 44 recommended conditions rescued 10 of 11 historically failed substrate pairs; high-confidence recommendations achieved a 33% success rate versus a 19% baseline.
Sources (1)
Nat. Comput. Sci. | Buchwald–Hartwig反应的稳健分布外预测
Public signals
OpenAlex citations 2
TL;DR - A Nature Computational Science study shows that standardized, chemically diverse data plus active learning enables robust out-of-distribution prediction of Buchwald–Hartwig reactions. The approach improves reaction-condition selection for unseen substrates without requiring complex models.
- Researchers combined 11,300 new automated experiments with public high-throughput data, creating a standardized dataset of about 27,500 reactions.
- Their diversity metric, CRDS, correlated more strongly with strict OOD performance than dataset size did (Pearson 0.79 versus 0.60).
- A random forest exceeded 90% ROC AUC on high-confidence OOD predictions, highlighting the importance of data quality and representation.
- Testing 44 recommended conditions rescued 10 of 11 historically failed substrate pairs; high-confidence recommendations achieved a 33% success rate versus a 19% baseline.