ASPECT - an intelligent approach to solvent selection
PDF Print E-mail

The design of an extensive polymorph screen demands careful consideration. How broad and diverse do the crystallization techniques need to be to understand the polymorph landscape? Which solvents or mixtures should be used? How can the number of experiments and cost be minimised while still having full confidence in the results?

One of the most important properties to consider when designing a solid form screen is the solubility of the compound in various organic solvents. It is clearly impractical to test the solubility in all the possible organic solvents and the thousands of mixtures of these solvents. At Crystallics we estimate the organic solubility using an empirical model for solubility with a NRTL-SAC model (see pictures 1 and 2).

Picture 1: Solvents and API have parameters X (hydrophobic), Y- and Y+ (polar), and Z (hydrophilic)

Picture 2: Solvent diversity from descriptor modelling

For each API a model is developed using experimental solubility data in 10 to 20 diverse organic solvents (see picture 3). These experiments are part of Crystallics’ enabling polymorph screens and not only help identify a suitable solid form for early development but also forms the basis for the design of extensive polymorph investigations later in development.

Picture 3: The NRTL-SAC empirical model used to estimate the organic solubility

With the combined solubility data from both the measurements and the modelling efforts, an extensive polymorph screen is designed. For each crystallization technique a different set of solvents must be selected to optimise the outcome.  For instance, the driving force for cooling crystallization comes from the temperature dependent solubility.  Cooling a saturated solution will give solid material. The yield Y of a cooling crystallization experiment is given by Y=VDC, where V is the volume and DC is the change in solubility (see picture 4). Only solvents that give sufficient yield will be selected for cooling crystallization technique.

Picture 4

An anti-solvent crystallization is based on reducing the solubility of the API in a solvent by the addition of another solvent.  Again, in this type of crystallization the yield is given by Y=VDC. The DC value in this case is the solubility decrease in the solvent/anti-solvent mixture (see picture 5). Only if there is less-than-linear solubility in the mixture (below the dotted line) will there be super-saturation and yield. The solvent/anti-solvent combinations that give sufficient yield will be selected in the anti-solvent screen design.

Picture 5

In vapor diffusion crystallization (whether it is onto amorphous solids or into saturated solutions) you need solvents with sufficiently high vapor pressure. In these crystallization experiments, the selected solvents have a minimal vapor pressure to ensure sufficient vapor-phase transport of the solvent onto the solid material or into the saturated solution.

To minimize the time it takes to select solvents for each experiment and to have confidence in finding as many forms as possible, Crystallics has developed a solvent selection informatics called ASPECT. This software allows the user to quickly select a diverse set of solvents and solvent mixtures, while applying constraints based on crystallization yield, boiling point, vapour pressure, solvent desirability (ICH class), solvent functional groups, et cetera. Once developed, the NRTL-SAC solubility model is integrated in the ASPECT program. It predicts the solubility at different temperatures and in solvent/anti-solvent mixtures while selecting the solvents, and applying the selection rules for each different technique. Solvents and mixtures that have the right solubility are selected by maximizing diversity in the selection, using molecular modelling and chemometric selection algorithm.