AbstractBACKGROUNDGlioblastoma (GB) is an aggressive brain cancer with poor prognosis and limited effective treatment. Many challenges hinder the success of anti-cancer therapies such as drug resistance, low surgical and pharmacological accessibility, and migration of cancer cell to nearing brain tissue. The origin of the tumoral population is often evasive due to a high tumoral cell heterogeneity, impeding drug discovery strategies.MATERIAL AND METHODSThis study uses the public scRNAseq atlas ExtendedGBMap (Ruiz-Moreno et al., 2022). The processed dataset was filtered to only keep 48 patient samples that contributed both tumor and normal cells, and only 8 cell types were selected, consisting of 4 tumor/healthy pairs: Astrocyte-like/Astrocyte cells, (50335/201 cells), Mesenchymal-like/Mural cells (32795/2528 cells), Neuron Progenitor-like/Neuron cells (8580/1091 cells) and Oligodendritic Progenitor-like/Oligodendritic Progenitor cells (38684/933 cells). All other cells were excluded. Patient-wise differential expression analysis was performed comparing tumor versus healthy cells, independently of cell-type, using scanpy.tl.rank_genes_groups default t-test, and patient were clustered based on their statistical test scores for each gene in the matrix using the k-means algorithm. Genes were ranked based on their statistical score, and gene set enrichment analysis was carried out using the WikiPathways, GO:MolecularFunction, Reactome, and Hallmark databases.RESULTSThis novel scRNAseq and paired tumor-healthy comparison based approach unraveled 3 main GB groups with distinct transcriptomic profiles. Group1 was enriched in DNA replication, glycolytic metabolism, amino-acid production, and Hypoxia. Group2 exhibited known activities of epithelial-to-mesenchymal transition, miRNA in cancer, and chondroitin/dermatan sulfate biosynthesis. Group3 was comprised of patients that did not seem to belong to any of the 2 groups, but that did not share any specific gene signature.CONCLUSIONWe present a new approach to GB patient subgroup identification with a potential for target discovery, exploiting bystander normal cells in high-throughput scRNAseq. This method partially overcomes inter-patient and intra-patient tumor heterogeneity and may support new therapeutic avenues for GB patients. We aim at validating and enriching this approach with paired scRNAseq and Spatial Transcriptomic data from the MOSAIC consortium.