Abstract: Anti-drug antibodies developed against therapeutic proteins have been shown to reduce the medication efficacy, and in worst case this immunogenicity can have safety implications for the patient. Understanding and predicting the immunogenic potential of therapeutic proteins has been a major challenge in the process of biotherapeutic drug discovery. In recent years, many AI methods have been applied to protein fragments to evaluate their immunogenicity potential and have shown encouraging accuracies. This presentation will provide an overview of popular AI techniques in this space, including natural language processing, position-specific scoring matrix, deep motif deconvolution, etc. Current use cases of such tools, research gaps, and future opportunities will also be discussed.
Bio: Jiayi Cox is a data scientist who delivers deep learning solutions on biologics at Novartis Institutes for BioMedical Research (NIBR). Her research interests include using graph neural network to model protein interactions and using language models to find protein binding site. Jiayi is experienced in several fields of studies including human genetics, molecular biology, and NGS data analysis. Prior to NIBR, she helped prioritize biomarkers nomination as a machine-learning scientist at a Boston-based pharmaceutical start-up. She obtained her PhD degree from Boston University on computational biology.