ML-COF: Covalent Organic Framework Synthesis Data with Success and Failure Records

A 2026 dataset by Bing Ma integrates 1822 in-house failed synthesis records and 2603 successful literature cases for covalent organic frameworks. It supports a random forest model achieving 91% solvent prediction accuracy and high R² scores for temperature and time prediction. The dataset underpins the ML-COF toolkit for predicting synthesis conditions.

Use Cases

• Training ML models for solvent prediction in COF synthesis based on integrated success-failure data.
• Predicting reaction temperature and time for novel COF frameworks based on the described model performance.
• Automating retrieval of known COF synthesis routes using the ML-COF toolkit mentioned in the description.
• Validating synthesis predictions for structurally complex COFs, such as M(salen)-COFs and M(Rabson)-COFs.

Strengths

• Integrates 1822 failed synthesis records, providing negative data often absent in materials datasets.
• Includes 2603 successful synthesis cases extracted from the literature.
• The derived model achieved 91% accuracy in solvent prediction and high R² scores (0.9129 for temperature, 0.9596 for time).
• Experimental validation successfully synthesized eight structurally complex COFs under predicted conditions.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is small in scale at 185.5 KB, indicating limited scope.

Provenance

Source

figshare

Collection Method

Curated from in-house laboratory records and extracted from literature.

Freshness

Last updated 2026-05-21 09:21:09; freshness should be verified.