Skip to main content
Fig. 1 | Acta Neuropathologica Communications

Fig. 1

From: Capture at the single cell level of metabolic modules distinguishing aggressive and indolent glioblastoma cells

Fig. 1

Spontaneous grouping of cancer cells by tumor of origin following unsupervised analysis. a Analytical and experimental strategy outline. b Normal cells group independently from tumor of origin. PCA (top) and chord (bottom) plots. Each dot represents a cell in PCA. b1: cells colored by normal cell type identity (purple: astrocytes; blue: oligodendrocytes; light blue: oligodendrocyte precursor cells; red: neurons; gold: myeloid cells; brown: vascular cells). Normal cell types determined as described [14]. b2: cells colored by tumor of origin (pink, green, orange, black for GBM1, 2, 4 and 6, respectively). c Cancer cells group by their tumor of origin. PCA (top) and chord (bottom) plots. Cells colored by tumor of origin (pink, green, orange, black for GBM1, 2, 4 and 6, respectively). d Impact of data treatment on the dependence of cell clustering to tumors. NMI: Normalized Mutual Information score. C: cells. MCH: metacell defined by hierarchical clustering. MCS: metacell defined by SNN (shared nearest neighbor) clustering. HKG: housekeeping genes. CNV: copy number variations. DE: differentially expressed. ODG: overdispersed genes. Black and white dotted lines: reference NMI scores of grouping analyses performed with all genes detected in GBM and normal cells, respectively. Note that NMI scores of GBM cell grouping remain constant, regardless of data normalization or filtering modes. Only data standardization reduces NMI score to a value similar to that obtained when analyzing normal cells. e Unsupervised analysis of data standardized by tumor results in clusters mixing cells from different tumors. PCA plots highlighting the tumor from which the cells derive (top: pink, green, orange, black for GBM1, 2, 4 and 6, respectively) or the 7 clusters identified (bottom) . f Gene ontology analysis of the genes describing each of the 7 clusters highlights a variety of biological processes, not linkable to specific functioning states. DAVID toolkit. Corresponding cluster number is indicated (colored as cluster colors in the bottom panel of e)

Back to article page