The UNIBI Team and its Expertise for CancerScan

The UNIBI team participating in the CancerScan project is part of the Genome Data Science (GDS) group at Bielefeld University, Germany. Led by Prof. Alexander Schönhuth, the GDS group specialises in biomedical data analysis using machine learning and data science methods. The team brings extensive experience from previous international projects, including pangenome research and studies on diseases such as ALS, COVID-19, and cancer.

Currently, Johannes Schlüter, a PhD student at GDS, is actively involved in CancerScan. His research focuses on machine learning methods for gene expression data in cancer, making his expertise highly relevant for the project. He will soon be joined by Raghuram Dandinasivara, also a PhD student at GDS, who contributes broad technical experience in computational biology and modern data processing technologies. As an expert for generative AI models in biomolecular systems PhD student Luna Pianesi provides her knowledge for the project.

Additionally, Kristin Willms, a current master student at Uni Bielefeld, collaborates with GDS and is currently occupied with her Master Thesis about cell-cell-communication networks.

Within CancerScan, the UNIBI team applies advanced machine learning and computational biology methods to analyse complex biomedical data. During the initial phase of the project, all partners collaboratively developed a comprehensive data model to integrate clinical, experimental, and analytical data into a unified knowledge graph. This interdisciplinary collaboration between clinicians, data scientists, and computational biologists has already laid the foundation for future analyses.

At the same time, the GDS group has started implementing first approaches to model cell–cell communication networks. Communication between cells plays a critical role in tissue function, and in cancer these communication patterns differ substantially from those in healthy tissue. Modelling these interactions is essential for building digital twins of cancer cell systems but poses significant computational challenges due to their complexity.

To address these challenges, the UNIBI team leverages modern machine learning approaches, in particular Transformer-based architectures. Originally developed for natural language processing, Transformers are well suited to capture complex contextual relationships. In CancerScan, cell–cell communication is treated as a structured language, allowing the model to identify patterns and assess the contextual importance of individual signalling components.

These initial developments mark an important step toward a deeper understanding of cancer cell communication. By combining machine learning with biological insight, the UNIBI team contributes to the long-term goal of creating accurate digital twins of cancer systems, supporting improved analysis and understanding of tumour biology.

 

Keywords

cell-cell-communication, attention, transformer, machine learning, embeddings, data analysis, genome data science, cancer research, digital twin