CRS 2024 Annual Meeting & Exposition

Molecular modeling and machine learning for small molecule and biologic drug formulation

Given the competitive market and inherent challenges in drug formulation, selecting and combining the right ingredients in the appropriate manner is essential. With advances in machine learning, physics-based simulation and compute hardware, modeling is emerging as a valuable source of information and knowledge to complement experimental characterization.

In this talk, we showcase several case studies illustrating how the tools from the Schrödinger platform can be applied to modeling formulations of small molecule and biologic drugs. Starting from data-driven approaches, we demonstrate how our machine-learning tools can provide an efficient prediction of drug solubility and other important properties of multi-component mixtures. We then describe physics-based approaches to study those complex and evolving structures, often in fluid states, that play a crucial role in the pharmaceutical industry.

For both small molecule and biologics formulations, we have developed powerful simulation tools employing atomistic or coarse-grained models to enable the characterization of molecular interactions and nanoscale structuring. For example, we can address the dissolution of amorphous solid dispersions, the self-assembly of polymer-based structures and the viscosity and aggregation of protein-excipient mixtures. We also present recent work on simulating the self-assembly and calculating the apparent pKa values of lipid nanoparticles used in the delivery of mRNA.