(448d) Automated Microfluidic Systems for Real-Time Pharmaceutical Crystal Metrology and Classification

Molecular crystals are ubiquitous in a variety of industrial contexts, from foods to chemicals and pharmaceuticals. The ability to characterize and control crystal attributes during crystallization, such as crystal size distribution (CSD) and crystal form, is crucial for ensuring in vivo efficacy of the final drug product. However, the CSD measurement and form characterization instruments available are either single point measurements or involve time lag between sampling and analysis respectively. This means that the kinetic parameters estimated, or crystal forms analyzed may differ depending on the time and/or spatial location of analysis. In this presentation, we present two methodologies involving microfluidic flow cells that enable the well-controlled spatial and temporal crystallization conditions, and when coupled with polarized light microscopy or image-based analytics, enables rapid, direct measurements and analysis of crystal attributes such as CSD and crystal forms at single crystal level.

We first demonstrate a well-based microfluidic flow cell platform for rapid and direct measurements of evolving crystal sizes and size-dependent growth rates within crystal ensembles exposed to well-defined flow fields. We present detailed growth measurements of two high aspect ratio model drugs - celecoxib and glycine, where growing crystal ensembles are trapped in pseudo-static fashion within wells in a microfluidic flow cell under controlled laminar shear fields and observed via polarized microscopy over sustained time intervals. Time-varying CSDs are extracted via an image segmentation-based length detection algorithm (ISLDA), which then allow rapid estimation of shear and size-dependent growth rates through a simple Eulerian-based optimization scheme. We also demonstrate discrimination between different regimes of crystal growth in the two model drugs. Next, we introduce a new methodology which utilizes a combination of microfluidic flow cells and a rotating polarizer-analyzer pair with orthogonally aligned polarization axes for automated access to interference colours of birefringent molecular crystals that are characteristic of the polymorphic form. When coupled with machine learning, this methodology enables unprecedented real-time and in situ classification of crystal ensembles (~3000 crystals classified in under 10 s) at a single crystal level. We demonstrate ~94% and ~86% accuracy in classification of two model systems comprising of polymorphs (α-glycine and β-glycine) and hydrates (azithromycin sesquihydrate and azithromycin dihydrate) respectively. We then apply the method to monitor dynamic transformation of molecular crystals from one form to another over time in crystal ensembles, through simultaneous form classification of crystals and direct crystal area measurements. This sheds quantitative insight into the dominant crystallization phenomena such as nucleation, growth or dissolution. The flow cells can also be retrofitted to crystallization vessels for rapid online measurements of evolving crystal sizes and forms.

We envision the applicability of these methodologies for crystallization kinetics modelling and in accelerating the exploration of storage, process condition or additive dependent polymorphic form outcomes, that are of interest during early-stage research and development when limited quantities of materials are available.