AI-Guided Discovery of New Targets for Cancer Immunotherapy

Cancer immunotherapy has transformed the treatment of several cancers, but its success depends on identifying the right targets. T cells recognise cancer through short peptides displayed on the surface of tumour cells by Human Leukocyte Antigen (HLA) molecules. These HLA-bound peptides provide a direct molecular snapshot of what is happening inside a cancer cell and represent attractive targets for cancer vaccines, T-cell receptor therapies and other precision immunotherapies. Modern immunopeptidomics and mass spectrometry can identify tens of thousands of HLA-presented peptides from tumours. Importantly, many potentially valuable antigens arise from regions of the genome and proteome that are missed by conventional protein databases, including non-canonical open reading frames, alternative translation products and other components of the “dark proteome”. Non-classical HLA molecules such as HLA-E and HLA-G may also present shared antigens that could enable broadly applicable, off-the-shelf immunotherapies. The major challenge is therefore no longer simply detecting peptides, but determining which of thousands of candidates represent the best therapeutic targets. This PhD project will develop artificial intelligence and computational approaches to identify and prioritise tumour antigens for cancer immunotherapy. The student will work with large-scale immunopeptidomics datasets generated from cancer cell lines, patient tumours and clinically relevant models, and integrate these with genomics, transcriptomics, proteomics, ribosome profiling and healthy-tissue datasets. A major focus will be the development of computational models that can distinguish therapeutically valuable tumour antigens from the much larger background of normal HLA-presented peptides. Candidate targets may include conventional tumour antigens, neoantigens, cryptic and non-canonical peptides, and peptides presented by both classical and non-classical HLA molecules. The project will address questions such as whether machine learning can predict tumour specificity, identify antigens shared across patients, predict which peptides are consistently presented at high abundance, and integrate multiple biological features to rank candidates for therapeutic development. Modern approaches including deep learning, protein language models and representation learning may be explored to identify patterns that conventional prediction algorithms miss. The long-term goal is to develop an AI-driven antigen discovery and prioritisation platform that can transform large multidimensional datasets into a short list of high-confidence therapeutic targets. This could substantially accelerate the development of cancer vaccines and off-the-shelf T-cell-based immunotherapies. Importantly, this will not be a purely computational project. The student will work closely with experimental scientists generating immunopeptidomics and functional data, allowing computational predictions to be experimentally tested and the resulting data to feed back into model development. What will the student learn? The successful candidate will receive training at the interface of artificial intelligence, computational biology, cancer immunology and translational research. Depending on their background and interests, they will develop expertise in machine learning and deep learning, protein language models, bioinformatics, multi-omics data integration, immunopeptidomics and mass spectrometry data analysis, HLA biology, antigen processing and presentation, cancer genomics and transcriptomics, and the discovery of canonical, cryptic and non-canonical tumour antigens. The student will also gain experience in developing reproducible computational pipelines, analysing large biomedical datasets, working with experimental scientists, and translating computational discoveries towards cancer vaccines and T-cell-based therapies. This project would particularly suit a student with a background in bioinformatics, computer science, artificial intelligence, computational biology, mathematics, data science, biomedical engineering or a related discipline who is interested in applying advanced computational methods to an important problem in cancer medicine. The bigger vision is to make tumour antigen discovery faster, more systematic and more predictive, enabling the development of precision and off-the-shelf immunotherapies that can benefit larger groups of cancer patients.