Protein biophysical properties as well as chemical modification are major concerns in the development of protein therapeutics. They affect biologics therapeutics' efficacy, manufacturability, and safety. We are interested in developing and implementing prediction methods to forecast the potential liabilities in biologics, including thermostability, aggregation, viscosity, deamidation, isomerization etc.<br/><br/>Due to the limited size of data sets and complicated structural features of proteins, prediction accuracy of those models is challenging. More accurate prediction methods would allow elimination or reduction of problematic variants as early as possible in the drug discovery process. We try to develop accurate prediction models by mining the key features of proteins (feature engineering) and applying machine learning techniques. Features includes in silico biophysical properties, local and global 3D structural information, especially the surface properties. Machine learning algorithms are able to select key features which contribute most to the predictive models through feature selection. Accurate prediction models can be derived by learning from the key features. Those features can also guide protein engineers for better design of new variants.

Prediction of protein biophysical properties and chemical modifications by correlating in silico protein structural features

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.