Transforming Solvent Selection in Crystallization Development with SoluAI; an AI-enabled Solubility Prediction Tool

Accurately predicting solubility remains one of the most persistent challenges in pharmaceutical development directly impacting bioavailability, drug substance process development, formulation strategy, and ultimately clinical success.  

Traditional approaches each come with trade-offs;  

1. Empirical methods require extensive lab work and time  

2. Semi-empirical models (e.g. QSPR) depend on existing data and can lack generalisability  

3. Theoretical approaches (e.g. COSMO-RS, MD) offer deep insight but can be computationally demanding and require calibration  

At TMA we have developed a hybrid AI enabled solubility prediction tool SoluAI, combining thermodynamic insights with advanced neural network technology. Trained on an extensive dataset covering more than 200 model compounds and 120 organic solvents, including solvent mixtures, the platform can rapidly predict temperature-dependent solubility across a broad range of solvent systems. 

SoluAI enables scientists to generate solubility-temperature profiles in seconds, supporting rapid virtual screening of solvent candidates for crystallisation processes such as cooling and antisolvent crystallisation. This allows teams to identify promising solvent systems earlier in development, reduce experimental burden, and focus laboratory resources where they deliver the greatest value. 

The results are compelling. Compared with leading open-source benchmarking tools, the model delivers significantly improved predictive performance, enabling more confident decision-making and more efficient process design. 

Key benefits of SoluAI

• Reduces experimental workload and costs associated with solvent selection  

• Supports more sustainable solvent selection 

• Creates tailored neural networks in minutes as new development data become available for prediction refinement  

• Delivers superior predictive performance compared with leading open-source alternatives 

  This approach empowers crystallization development teams to accelerate development timelines, support sustainability goals through smarter solvent selection, and unlock a more agile, data-driven path to process optimisation. 

Recent publications

• Mac Fhionnlaoich, N., Zeglinski, J., Simon, M., Wood, B., Davin, S., & Glennon, B. (2024). A hybrid approach to aqueous solubility prediction using COSMO-RS and machine learning, Chemical Engineering Research and Design.  https://doi.org/10.1016/j.cherd.2024.07.050 

• Cheemarla, V., Mac Fhionnlaoich, N., Wood, B., Davin, S., & Glennon, B.  
  SoluAI – AI-enabled solubility modelling for crystallisation process design.  23rd International Symposium on Industrial Crystallization, 6 -9 September 2026, Budapest, Hungary. https://akcongress.com/isic2026/