Computational methods for optimized biologics formulation

点击翻译按钮获取中文摘要

Project Summary: Protein-based biologics – therapeutics whose active ingredient is a protein and most commonly a monoclonal antibody (mAb) – make up a $200 billion/year market that is expected to double in size by 2025. A critical component in the safety and efficacy of biologics is the need to maintain the active protein during long-term storage and subsequent injection/infusion. The selection of excipients and buffers toward this end is termed “formulation.” Proper formulation of a protein-based drug is essential to stabilize the active protein from unfolding and to block sites on the folded protein that may pose an aggregation risk and lead to elevated viscosity due to undesirable protein-protein interactions (PPI). Importantly, formulation can be done without altering the sequence of (i.e. re-engineering) the protein and is thus an independent tool for bringing a biologic therapeutic to market. Current approaches to choosing an optimized formulation are either low-throughput experiments or computational methods that do not take into account the molecular details of excipient-protein interactions. The established Site Identification by Ligand Competitive Saturation (SILCS) computational platform technology maps the affinity pattern of the complete 3D surface of a protein for a wide diversity of chemical functional groups. The functional group affinity pattern is then used to determine excipients that can bind to and stabilize the active, folded conformation of a protein and bind to regions of the protein that may participate in PPI, thereby inhibiting aggregation and decreasing viscosity. The broad goal of the proposal is the continued development and validation of SILCS-Biologics as an industry-ready workflow and a graphical user interface to manage and apply the extensive information generated by SILCS excipient screening and PPI analysis. In the proposed studies experimental efforts will be undertaken to generate data for model training and validation across a variety of proteins and commonly used excipients. That data will be then combined with computed SILCS metrics including excipient binding locations and affinities and potential regions that can participate in PPI across the complete protein surface, including the impact of pH and the unique properties of Arginine as an excipient. This information will then be applied in the context of machine learning to develop models that will predict excipients that will block PPI thereby lowering viscosity and slowing aggregation as well as stabilize the folded, biologically active state of the protein. The proposed models will be validated at the University of Maryland, Baltimore and with industrial and government partners against established experimental methods on a range of proteins with various therapeutic indications. Upon successful completion of the project new offerings will be added to the existing SILCS software suite that will minimize the time and costs requirements for the formulation of biologics as well as lead to improved formulations thereby improving clinical outcomes. These capabilities will be implemented in the context of industry-ready workflows for direct sale to pharmaceutical companies and for use in contract research for the optimized formulation of biologics.