Practical Data-Driven Interrogation of Reactivity in Acid-Catalyzed Carbonyl–Olefin Metathesis with Machine Learning and Large Language Models
Bibliographic record
Abstract
Carbonyl–olefin metathesis (COM) has emerged as a powerful yet mechanistically complex transformation for forging carbon–carbon bonds. Although diverse Brønsted and Lewis acid catalysts enable COM reactivity, predicting which catalyst will be effective for a given substrate remains challenging. Likewise, machine learning (ML) and large language models (LLMs) are emerging tools for planning experiments in organic synthesis, but their relative strengths in mechanistically rich catalytic systems like COM remain unclear. Here we address both challenges through a practical ML framework that combines a rigorously curated, blinded dataset of 147 COM reactions—spanning structurally diverse substrates and Brønsted, Lewis, and superelectrophilic ion-pair catalysts—with tiered predictive models to guide catalyst and substrate selection. A Morgan-fingerprint baseline captures substantial yield variation, demonstrating that structure-only representations provide rapid initial screening. Physically informed models based on quantum-derived descriptors of independent catalyst and substrate structures achieve higher accuracy (R² ≈ 0.92) and maintain performance on external substrates, while feature-importance and SHAP analyses reveal catalyst HOMO energy, dimerization propensity, and substrate carbonyl polarization as dominant determinants of reactivity. Reactivity-cliff analysis distinguishes smoothly interpolating catalysts such as FeCl₃ and ZnCl₂ from cliff-rich ion-pair systems, identifying regions where predictions are intrinsically unreliable and targeted experiments are most informative. Variance-guided Bayesian optimization recovers most full-model performance using only ~40% of the reaction matrix, while a parallel GPT-4–driven, code-free acquisition protocol surpasses expert selection and substantially outperforms random sampling. Together, these results provide a mechanistically interpretable, data-efficient blueprint for COM catalyst discovery and demonstrate how uncertainty-aware ML and LLMs can augment human intuition in complex catalytic systems.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".