Design of Spherical Crystallization of Active Pharmaceutical Ingredients via a Highly Efficient Strategy: From Screening to Preparation

This work aims to develop a highly efficient spherical crystallization from screening to preparation stage based on liquid-liquid phase separation (LLPS). Mixtures than can undergo an LLPS split into two liquid phases with different physical properties, and the oil droplets formed during that process make LLPS a promising approach to prepare spherical particles of an active pharmaceutical ingredient (API). In the screening stage, three machine learning (ML) models (artificial neural network, support vector machine, and logistic regression) were established for predicting LLPS for an API. Two linear models, a simple linear model and a machine learning-based linear model, were also constructed to produce further optimization. The ML-based prediction of LLPS was first established in this work and showed high accuracy and reliability. Also, when compared to a method where the screening depended on the results of experiments, the prediction model highly reduced the use of chemical substances and saved labor and time. In the preparation stage, water and ethanol, which have low toxicity to mammals and have environmental advantages over other organic solvents, were applied as the solvents of LLPS-based spherical crystallization. The LLPS-based preparation process of spherical particles possesses advantages in terms of reduction of the number of unit operations as well as energy consumption and processing cost.

Automated summary

This work aims to develop a highly efficient spherical crystallization process based on liquid-liquid phase separation (LLPS) from screening to preparation stage. Machine learning models were used to predict LLPS for an API, with water and ethanol chosen as solvents to reduce toxicity and energy consumption.