Search

We present deep learning to predict the flowability of pharmaceuticals from microscopy images. This enables flowability assessments with smaller API quantities, saving experiment time and costs when material is limited during early drug development.

A generic framework for enhancing an initial solubility prediction with ML, even with simple methods and a modestly sized, sparse dataset. We dissect the setup to show the model “locking on” to the target system as more data are made available.

A predictive tool was developed to aid process design and to rationally select optimal solvents for isolation of active pharmaceutical ingredients. The objective was to minimize the experimental work required to design a purification process by (i) starting from a rationally selected crystallization solvent based on maximizing yield and minimizing solvent consumption (with the constraint of maintaining a suspension density which allows crystal suspension); (ii) for the crystallization solvent identified from step 1, a list of potential isolation solvents (selected based on a series of constraints) is ranked, based on thermodynamic consideration of yield and predicted purity using a mass balance model; and (iii) the most promising of the predicted combinations is verified experimentally, and the process conditions are adjusted to maximize impurity removal and maximize yield, taking into account mass transport and kinetic considerations. Here, we present a solvent selection workflow based on logical solvent ranking supported by solubility predictions, coupled with digital tools to transfer material property information between operations to predict the optimal purification strategy. This approach addresses isolation, preserving the particle attributes generated during crystallization, taking account of the risks of product precipitation and particle dissolution during washing, and the selection of solvents, which are favorable for drying.

Silk continues to amaze: over the past decade, new research threads have emerged that include the use of silk fibroin for advanced pharmaceutics, including its suitability for drug delivery. Despite this ongoing interest, the details of silk fibroin structures and their subsequent drug interactions at the molecular level remain elusive, primarily because of the difficulties encountered in modeling the silk fibroin molecule. Here, we generated an atomistic silk model containing amorphous and crystalline regions. We then exploited advanced well-tempered metadynamics simulations to generate molecular conformations that we subsequently exposed to classical molecular dynamics simulations to monitor both drug binding and release. Overall, this study demonstrated the importance of the silk fibroin primary sequence, electrostatic interactions, hydrogen bonding, and higher-order conformation in the processes of drug binding and release. �� Copyright �� 2019 American Chemical Society.