AliGhiami/Solubility

fastsolv — Organic Solubility Predictor

This Space wraps the fastsolv model by Attia, Burns, Doyle & Green (MIT, 2025), published in Nature Communications (doi:10.1038/s41467-025-62717-7).

What it does

Given a solute (SMILES), a solvent (predefined or custom SMILES), and a temperature range (K), it predicts:

• —Predicted logS — base-10 log of molar solubility (mol/L)
• —Standard deviation — uncertainty from a 4-model ensemble

Model details

• —Trained on BigSolDB: 54,273 measurements, 839 solutes, 138 solvents
• —Temperature range: −30 °C to 130 °C (243 K – 403 K)
• —Architecture: fastprop descriptor-based neural network ensemble

Citation

Attia, L., Burns, J. W., Doyle, P. S., & Green, W. H. (2025).
Data-driven organic solubility prediction at the limit of aleatoric uncertainty.
Nature Communications, 16(1), 7497.