Background
Melanoma, the most aggressive form of skin cancer, represents a major clinical challenge due to its highly variable progression, immune responses and therapeutic outcomes. Despite being the archetypal immunogenic tumour, there is currently no reliable means of predicting prognosis or patient response to immunotherapy. Advances in molecular profiling at single-cell and spatial resolution have created new opportunities to dissect tumour–immune interactions [1, 2, 3]. However, the integration of diverse datasets into predictive models of disease course remains an unmet need.
Novelty & Importance
This project will pioneer computational approaches to delineate immune signalling in solid tumours by integrating multi-omic, single-cell and clinical data. We will employ advanced machine learning (ML) and artificial intelligence (AI) approaches, including knowledge graphs, deep neural networks, and optimisation pipelines to create predictive models of tumour–immune dynamics [4, 5, 6]. Unlike existing studies that focus primarily on descriptive profiling, this project aims to deliver clinically relevant tools that can stratify patients, identify prognostic immune features, and support precision immunotherapy. By bridging computational innovation and cancer immunology, this work has the potential to generate both methodological advances in data science and tangible improvements in cancer patient care.
Aims & Objectives
The project will:
• Develop interpretable AI/ML frameworks for immune cell type annotation, signalling pathway analysis, and immunotherapy response prediction.
• Construct high-resolution immune repertoire atlases. capturing clonotypes, signalling networks, and cellular interactions.
• Disseminate novel computational tools and data resources as open-access platforms to benefit both the bioinformatics and oncology communities.
Through this integrative approach, the project aims to advance our understanding of immune mechanisms in cancer progression and contribute to the development of predictive tools for personalised immunotherapy.
References
1. da Costa Avelar PH, Laddach R, Karagiannis SN, Wu M, Tsoka S (2023). Multi-omic Data Integration and Feature Selection for Survival-Based Patient Stratification via Supervised Concrete Autoencoders. In: Nicosia, G., et al. Machine Learning, Optimization, and Data Science. LOD 2022. Lecture Notes in Computer Science, vol 13811. Springer, Cham. https://doi.org/10.1007/978-3-031-25891-6_5
2. Crescioli S, Correa I, Ng J, Willsmore ZN, et al, Karagiannis SN (2023). B cell profiles, antibody repertoire and reactivity reveal dysregulated responses with autoimmune features in melanoma. Nat Commun. 14(1):3378. https://doi.org/10.1038/s41467-023-39042-y
3. Amiri Souri E, Chenoweth A, Cheung A, Karagiannis SN, Tsoka S (2021). Cancer Grade Model: a multi-gene machine learning-based risk classification for improving prognosis in breast cancer. Br J Cancer 125, 748–758. https://doi.org/10.1038/s41416-021-01455-1
4. Chen Y, Liu S, Papageorgiou LG, Theofilatos K, Tsoka S (2023). Optimisation Models for Pathway Activity Inference in Cancer. Cancers. 2023; 15(6):1787. https://doi.org/10.3390/cancers15061787
5. Liapis GI, Tsoka S, Papageorgiou LG (2025). Optimisation-Based Feature Selection for Regression Neural Networks Towards Explainability. Machine Learning and Knowledge Extraction, 7(2):33. https://doi.org/10.3390/make7020033
6. Ge S, Sun S, Xu H, Cheng Q, Ren Z (2025). Brief Bioinform, bbaf136. (doi 10.1093/bib/bbaf136)