A Multi-Layer Kernel Framework for the Prediction of Binding Affinity Between Immunological Proteins and Antigenic Peptides
DescriptionIn this project, we propose a new machine learning framework for predicting the binding affinity between antigenic peptides and immunological proteins. For immunological proteins, we focus on the Major Histocompatibility Complex (MHC) class II molecules, which present antigenic peptide fragments to helper T cells for immune response initiation. This study focus is in view of the highly polymorphic nature of MHC II, and the scarcity of available experimental data to allow accurate binding characterization. Our main objectives are as follows:To design a multi-layer kernel representation for antigenic peptides and MHC II molecules, which takes into account the short and long range patterns of their amino acid residues.To design a predictor ensemble architecture for binding prediction, which can integrate different prediction models to improve the final result.To design a multi-objective optimization evolutionary algorithm to select the optimal predictors in the ensemble architecture.
|Effective start/end date||1/05/12 → 24/02/15|