(642b) Computer-Aided Design of Solvent Blends for the Cooling and Anti-Solvent Crystallisation of Active Pharmaceutical Ingredients

Watson, O. L. - Presenter, Imperial College London
Adjiman, C. S., Imperial College London
Galindo, A., Imperial College London
Jackson, G., Imperial College London
The majority of pharmaceutical products are delivered in the solid form, such as tablets or via aerosols. As such, crystallisation remains the process of choice for the manufacturing of active pharmaceutical ingredients (APIs) – typically via solvent-based techniques. Thermodynamically, solvent choice will affect the solubility, and thus crystal yield of the API, but the effects of this decision on transport properties, process and product safety, and the final crystal morphology – which can impact downstream processing and ultimately the in vivo efficacy of the drug – are all driving forces for selection. For novel drug molecules, solvent selection is often performed via time-consuming and expensive experiments [1], limiting the full range of solvent mixtures and process conditions from being explored. Consequently, fast development of manufacturing processes that achieve high standards in terms of productivity, quality and environmental performance can be difficult.

Over the last decades, Computer Aided Molecular Design (CAMD) has been developed to overcome such difficulties, with the ultimate aim of guiding experiments towards optimal candidate molecules [2,3]. To reduce the number of potential solutions in such designs, a decomposition-based approach has been proposed [4], whereby smaller subproblems are posed and then solved successively. By fixing process conditions and utilising a single crystallisation technique, this has been applied to crystallisation design [5]. Indeed, most existing methodologies focus on the selection of single solvents, utilising a specific crystallisation technique, whereas more integrated problems, in which cooling and anti-solvent effects are combined and solvent mixtures are considered, have not yet received significant attention. This is likely due to the complexity of formulating and resolving a mixed-integer optimisation problem to represent these design choices; such problems result in challenging non-convex design spaces for the continuous variables, in addition to the combinatorial solution space.

More recently, the use of Generalised Disjunctive Programming (GDP) [6] within the Computer Aided Mixture/blend Design (CAMbD) framework has been proposed to design optimal solvent mixtures that maximise solubility of an API [7,8]; optimal mixtures were shown to outperform pure compounds. This strategy can be used to use to solve more integrated problems, enabling the user to explore the relative merits of cooling crystallisation, anti-solvent crystallisation and combined approaches concomitantly and solvent mixtures can be considered throughout [9], but this type of approach has not yet received significant attention. We investigate the formulation and solution of such an integrated crystallisation design problem, whereby the optimal process temperature, selection of solvent and anti-solvent molecules, and design of the solvent mixture composition, can be identified simultaneously – resulting in a mixed integer nonlinear problem (MINLP). Based on an equilibrium model, the formulation considers solvent use and crystal yield, but can be extended to account for other factors. To calculate the relevant thermodynamic properties, we use the SAFT-γ Mie equation of state, which has been shown to provide high quality predictions of solubility [10].

This design formulation is implemented in gPROMS and successfully applied to a case study on the crystallisation of lovastatin, showing that this more general approach to crystallisation design can be used effectively to optimise desired metrics, exceeding solutions that would otherwise be achieved by decomposition-based methods. Furthermore, it is shown that coupling of cooling and anti-solvent crystallisation can achieve greater crystal yields, whilst reducing the mass of solvent required for the process, further emphasising the usefulness of this approach, and of solvent mixtures in general.

[1] Brown CJ, McGlone T, Yerdelen S, Srirambhatla V, Mabbott F, Gurung R, et al. Enabling precision manufacturing of active pharmaceutical ingredients: workflow for seeded cooling continuous crystallisations. Molecular Systems Design & Engineering. 2018.

[2] Achenie L, Venkatasubramanian V, Gani R, editors. Computer aided molecular design: theory and practice. Elsevier. Amsterdam, The Netherlands; 2003.

[3] Gani R. Chemical product design: challenges and opportunities. Computers & Chemical Engineering. 2004;28(12):2441-57.

[4] Karunanithi AT, Achenie LE, Gani R. A new decomposition-based computer-aided molecular/mixture design methodology for the design of optimal solvents and solvent mixtures. Industrial & engineering chemistry research. 2005;44(13):4785-97.

[5] Karunanithi AT, Achenie LE, Gani R. A computer-aided molecular design framework for crystallization solvent design. Chemical Engineering Science. 2006;61(4):1247-1260

[6] Raman R, Grossmann IE. Modelling and computational techniques for logic based integer programming. Computers & Chemical Engineering. 1994;18(7):563-78.

[7] Jonuzaj S, Akula PT, Kleniati P, Adjiman CS. The formulation of optimal mixtures with generalized disjunctive programming: A solvent design case study. AIChE J. 2016;62(5):1616-1633

[8] Jonuzaj S, Adjiman CS. Designing optimal mixtures using generalized disjunctive programming: Hull relaxations. Chemical Engineering Science. 2017; 159 106-130.

[9] Watson OL, Galindo A, Jackson G, Adjiman CS. Computer-aided design of solvent blends for the cooling and anti-solvent crystallisation of ibuprofen. European Society of Computer-Aided Process Engineering. 2019, submitted

[10] Hutacharoen P, Dufal S, Papaioannou V, Shanker RM, Adjiman CS, Jackson G, Galindo A. Predicting the solvation of organic compounds in aqueous environments: from alkanes and alcohols to pharmaceuticals. Industrial & Engineering Chemistry Research. 2017;56(38):10856-76.