Colorectal cancer (CRC) is characterized by a complex and heterogeneous tumor microenvironment, where accurate identification of constituent cell types is essential for understanding disease progression and therapeutic response. Recent advances in spatial transcriptomics (ST) allow for high-resolution mapping of gene expression across tissue architecture, offering a powerful approach to dissect the spatial organization of cells in situ. In parallel, image-based analysis of histological slides, such as hematoxylin and eosin (H&E) stained sections, remains a widely used and accessible method for estimating cell type composition using morphological features.  This thesis aims to systematically compare cell type predictions derived from ST-based spatial deconvolution methods (e.g., Cell2Location) and from image analysis pipelines (e.g., MIKAIA software) across 14 publicly available CRC samples. By integrating and contrasting these methodologies, this study seeks to evaluate their concordance, highlight their respective strengths and limitations, and contribute to a more robust framework for cellular characterization in CRC tumors.

The student will:

1. Familiarize herself with theoretical background and necessary tools, in particular: background of cell type composition and histopathology of colorectal carcinoma tumours pipelines of spatial transcriptomics data analysis (quality control and normalization) MIKAIA sw for image analysis from H&E slides spatial cell deconvolution techniques such as Cell2Location
2. Perform spatial transcriptomics data analysis of 14 publicly available CRC samples including QC and normalization
3. Apply spatial cell type deconvolution and derive proportions of different cell types in the spots
4. Analyze cell type composition using MIKAIA sw on the respective H&E slides
5. Compare the predictions from spatial transcriptomics and image analysis and interpret the findings