It describes the successful application of a novel quantitative modeling technique (support vector machine regression, or SVR) in the modeling of an important biological process: the interaction between peptide epitopes and major histocompatibility complex proteins (MHCs). The models we have established outperformed existing models constructed using less sophisticated techniques in the field.

“The effectiveness of the SVR-based modeling of peptide-MHC interaction is another demonstration of learning-based techniques in solving real-world problems.”

It describes a methodology. SVR differs from "classical" quantitative modeling techniques in that it is a machine learning technique. The effectiveness of the SVR-based modeling of peptide-MHC interaction is another demonstration of learning-based techniques in solving real-world problems.

The T cell is a specialized type of immune cell continuously searching out proteins originating from pathogenic organisms. The T cell surface is enriched in a particular receptor protein: the T cell receptor or TCR, which binds to major histocompatibility complex proteins (MHCs) expressed on the surfaces of other cells. MHCs bind small peptide fragments derived from pathogen proteins. It is the recognition of such complexes that lies at the heart of the cellular immune response.

In this research, we applied the machine learning based SVR technique to model the binding affinities between MHC and small peptides, and achieved models with considerably better performance than models previously developed in the field.

This is collaborative research conducted with Dr. Darren Flower’s Bioinformatics group at Oxford. Dr. Flower is a leading expert in immunoinformatics, the application of bioinformatics to immunology. We are keen to transfer the latest developments in machine learning theories and applications and thus solve real-world life-science problems—so this project started quite naturally.

On the problem domain, improved models of peptide-MHC interactions will lead to a savings in cost and experimental effort in immunology research, and, in the long run, will improve people’s health. On the methods’ side, successful application of a recently developed learning-based technique in a real-world problem will help promote the development of more advanced machine learning methodologies.