375130 Method for Prediction of Pharmaceutical Solubility for Solvent Selection

Monday, November 17, 2014: 1:45 PM
203 (Hilton Atlanta)
Andrew Bird, Scale-up Systems, Dublin, Ireland and Joe Hannon, Scale-up Systems Inc, Wilmington, DE

A method is described for modeling the solubility of organic, non-electrolyte solutes in common solvents and mixtures.  This has applications for solvent selection for reaction, crystallization and extraction operations as well as in formulation design.  A small number of solubility measurements are required to fit model parameters; thereafter, solubility predictions can be made in a wide range of single, binary and ternary solvent systems (e.g. for identification of conditions for a solubility maximum in mixtures).

These calculations assist scientists and engineers in selecting solvent systems and ensuring that promising candidates that may not be ‘on the shelf’ are given due consideration.

The method comprises adding an extra group representing the solute molecule to the UNIFAC (ref 1) tables, including its UNIFAC R and Q parameters, then fitting the interaction parameters between the solute and the functional groups of the solvents used when measuring solubility.  An alternative tuning of standard UNIFAC has been described elsewhere (ref 2).  Heat of fusion and melting temperature are inputs (preferably) and can also be fitted.  Predictions can then be made of solubility in any solvent system where these functional groups are present.

The method overcomes some traditional weaknesses of UNIFAC while maintaining the standard table of parameters intact.  These include:

-          Application for solutes that are pharmaceutically relevant and cannot be fragmented into UNIFAC groups

-          Much improved accuracy compared to normal UNIFAC, which precludes regression of parameters.

Some recent applications of the method will be reviewed to indicate its performance relative to other techniques for solubility prediction (refs 3, 4 and 5).

Utilizing the functional group contribution approach (rather than conceptual segments) maintains advantages that may be leveraged in future development of the method to predict solubility of structurally similar impurities as well as primary solutes. 

The current implementation considers 163 solvents, enabling potential prediction in thousands of binary solvent mixtures and several million ternary solvent mixtures.  Users can limit the scope to solvents that are desirable for e.g. regulatory, cost or environmental considerations. 


  1. Fredenslund, A., Jones, R. L., Prausnitz, J. M., Group-Contribution Estimation of Activity Coefficients in Nonideal Liquid Mixtures. AIChE J., 1975. 21(6): p. 1086-1099
  2. Modarresi, H., Conte, E., Abildskov, J., Gani, R., & Crafts, P. A. (2008). Model-Based Calculation of Solid Solubility for Solvent SelectionA Review. Industrial & Engineering Chemistry Research, 47(15), 5234–5242. Retrieved from http://pubs.acs.org/doi/abs/10.1021/ie0716363
  3. E. Sirota and T. Lin, Modelling Small Molecule Synergistic Solubility Behavior, Presented at DynoChem User Meeting, Chicago, May 2011: available online at https://dcresources.scale-up.com/Default.aspx?id=388  
  4. N. Chennamsetty, O Lyngberg & J. Qiu, Determining Solubility of APIs and Intermediates From Automated Parallel Experiments and Modeling, AIChE Annual Meeting 2011, Paper 56c: http://www3.aiche.org/proceedings/Abstract.aspx?PaperID=236416
  5. New solubility prediction utility for solvent selection in DynoChem, Scale-up Systems blog, March 2010: http://blog.scale-up.com/2010/03/new-solubility-prediction-utility-for.html

Extended Abstract: File Not Uploaded