(53g) Modeling Minimum Free Energy Path for Crystallization of Drug Molecules in Different Solvents

Authors: 
Liu, C., North Carolina State University
Santiso, E. E., North Carolina State University
Wood, G., Pfizer
Crystallization from solution is fundamental to drug production in pharmaceutical industries where about 90% of the products contain one or more crystalline materials. As it is a combined result of nucleation and crystal growth, crystallization is affected by different precipitation conditions, such as solvents, presence of additives, etc.. Studying the mechanisms that how drug molecules aggregate and organize themselves to crystallize in different solvents is important to many aspects in drug design and production, including polymorph selection and nucleation rate. While such molecular motions are possible to be observed via computational simulations, it remains a challenge for standard molecular modeling with nucleation being a rare event. In addition, to characterize the transition from liquid to solid state requires an effective measurement of local molecule order which is not an easy task. In this work, we implemented the string method in collective variables in combination with the construction of order parameters to sketch a minimum free energy path for crystallization of drug molecules in different solvents. Two types of drug molecules, sulfadiazine and sulfamerazine (Form II), and three different solvents, water, methanol and acetonitrile, were selected.

Firstly, force field parameters for sulfadiazine and sulfamerazine were developed using the CHARMM general force field (CGenFF). The partial charges were modified to reproduce interactions with explicit TIP3P water molecules at quantum mechanical level of theory. The bond, angle and dihedral parameters were optimized by fitting them to yield a minimum energy geometry while capturing the potential energy surface calculated by quantum mechanics. The structural error of the unit cell parameters for sulfadiazine decreased from 3-8% before optimization to 0.3-1.5% after refitting the parameters. For sulfamerazine, the error decreased from 6.7-11.6% to 0.1-1.8%. Secondly, we constructed order parameters (OPs) from a generalized pair distribution function which gives the probability that a molecule with a particular internal configuration has a neighboring molecule with a given internal configuration, a given center of mass distance away from the first molecule and a given orientation with respect to the first molecule. The OPs of bond orientation and relative orientation have well separated distributions and both are able to quantitatively distinguish between liquid and solid states of the drug molecules. Finally, the minimum free energy paths were determined by using string method in collective variables while the constructed OPs parameterized the â??reaction coordinateâ? during the crystallization of sulfadiazine and sulfamerazine from their liquid states to solid states.