Benchmarking Cellular Deconvolution Methods for Spatial Transcriptomics Data

RNA sequencing technologies enable the simultaneous detection of tens of thousands of genes, providing researchers with a powerful tool for identifying novel genes involved in biological processes and potential targets for new treatments. A recent advancement in this field is spatial transcriptomics, which allows for the analysis of gene expression within the context of tissue architecture. The 10x Genomics Visium platform is a leading technology in this field, utilizing a specialized slide with barcoded spots to capture transcriptomic data from specific regions of a tissue section, allowing for detailed spatial mapping of gene expression. However, each spot contains multiple cells, which must be deconvolved to accurately identify the cell types present. This project aims to evaluate and compare various cellular deconvolution techniques to determine their effectiveness in analyzing Visium data. We will use Visium datasets with known cellular compositions and apply different deconvolution methods, including both traditional algorithms and advanced machine learning models. By assessing metrics such as accuracy, precision, and computational efficiency, we will compare the effectiveness of each method in reconstructing cellular compositions and spatial distributions.