147 Carbonyl-Olefin Metathesis Reactions with Catalyst Performance Data

A blinded dataset of 147 carbonyl–olefin metathesis reactions underpins a machine learning framework for predicting catalyst effectiveness. The dataset, curated by Zi-Hao Li and shared on figshare in 2026, includes quantum-derived descriptors and reaction outcomes. It supports models achieving high accuracy (R² ≈ 0.92) and identifies key reactivity drivers like catalyst HOMO energy.

Use Cases

• Training structure-based screening models based on Morgan fingerprints mentioned in the description
• Building high-accuracy predictive models for catalyst performance using quantum-derived descriptors
• Conducting feature-importance and SHAP analyses to identify dominant reactivity drivers like catalyst HOMO energy
• Performing reactivity-cliff analysis to distinguish smoothly interpolating catalysts from cliff-rich systems
• Guiding experimental prioritization using variance-guided Bayesian optimization or LLM-guided acquisition protocols

Strengths

• Contains 147 rigorously curated and blinded reactions, providing a controlled experimental matrix
• Supports models achieving high predictive accuracy (R² ≈ 0.92) with retained performance on external substrates
• Includes quantum-derived descriptors enabling mechanistic interpretation of reactivity drivers

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Dataset size is 318.4 KB, indicating a limited scope suitable for proof-of-concept studies

Provenance

Source

figshare

Collection Method

Likely contains experimental data from a curated study of carbonyl–olefin metathesis reactions.

Freshness

Last updated 2026-05-15 18:04:23; freshness should be verified