Cell-BiID

Intra-cancer and between cancer heterogeneity is a well known problem that limits the efficacy of targeted therapies. The characterization of this heterogeneity at different scale is studied using single-cell transcriptomic. Our aim is to analyse and compare scRNA-seq datasets of Glioblastoma and Medulloblastoma with Cell-ID, a multivariate statistical method for the extraction of per-cell gene signatures based on multiple correspondence analysis. More specifically, we want to explore whether it is possible to find overlapping communities of different cell-types between tumours. Using Cell-ID we are able to match cell-type labels between different datasets. Starting from this point we build a bipartite graph with two sets of nodes, one for each sample, and links weighted by P-values that measure the similarity of the cells. Finally we run community detection algorithms. We show that on this bipartite graph information about the two tumours considered is recovered and this enable us to relate similar samples as well as similar cell-types.