Fig. 8

Analysis of the single-cell composition of the four molecular subtypes of HGSTOC. A Venn diagram illustration the molecular subtype classification of the 7 patients included in this study, by four different subtyping algorithms using conventional bulk RNA sequencing data. Coloured dots indicate an individual label given to a particular patient. Dots positioned in the four-way intersection, indicate a unique label agreed upon by all subtyping algorithms (patients 1 and 2). Patients 4, 5, 6 and 7 received two different labels, patient 3 three different labels. Subtyping algorithms used were as follows: Helland et al. (Plos One 2011), Verhaak et al. (JCI 2013), Konecny et al. (JNCI 2014), ConsensusOV Chen et al. (CCR 2018). B Kaplan-Meier curves indicating the difference in survival time across the 1467 HGSTOC patients included in our reference cohort based on the ConsensusOV algorithm. Results of the log-rank test confirmed the prognostic value of the molecular subtypes (p < 0.001). Patients still alive at the time of analysis were censored at the time they were last followed up. Survival curves are unadjusted for covariates and the analysis includes all randomly assigned patients. C Molecular subtype scores of our 18,403 single cells calculated by the ConsensusOV package. For all 18,403 single cells the differential gene expression data were used to individually score each cell for the 4 molecular subtype signatures. Global scoring for all cells in one subcluster is visualised by boxplot presentation. Prognostic cell phenotypes were marked in bold. D Violin plots showing the enrichment scores of the four molecular ConsensusOV subtypes in each cell of the 6 prognostic subclusters