Melting Points for 139 Nonionic Deep Eutectic Solvents with Machine Learning Predictions

A computational chemistry dataset containing 1824 experimental melting point measurements across 139 binary nonionic deep eutectic solvent (DES) systems. Machine learning models, including an XGBoost model achieving an R² of 0.909, were developed to predict melting points using molecular and physicochemical descriptors. The dataset was authored by María A. Escobedo and last updated on May 14, 2026.

Use Cases

• Training machine learning models for melting point prediction based on molecular descriptors like PEOE_VSA1.
• Virtual screening of candidate DES formulations to reduce experimental burden based on the described computational framework.
• Analyzing the influence of hydrogen bond donor vs. acceptor properties on melting point behavior based on SHAP analysis results.
• Studying nonadditive interactions in solvent mixtures based on mixture-weighted descriptors accounting for 58% of feature importance.

Strengths

• Contains 1824 experimental measurements providing a substantial basis for model training.
• Model performance is quantitatively reported with R²=0.909, RMSE=17.8 K, and MAE=10.8 K.
• Feature importance is analyzed, revealing the dominance of charge-weighted molecular surface area (PEOE_VSA1) and pure component melting temperatures.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect bias inherent to the specific experimental sources aggregated on figshare.

Provenance

Source

figshare

Collection Method

Experimental measurements compiled for machine learning modeling.

Time Range

null

Freshness

Last updated 2026-05-14 06:32:04; freshness should be verified.

Geography

null