Browsing by Subject "random forest"

High-grade Serous Cancer (HGSC) stands out as the most prevalent and lethal subtype of ovarian cancer, characterized by genomic instability and aggressive behaviour. Genetic alterations are pivotal in its development and progression, which leads to distinct molecular profiles and clinical behaviours. Recent attention has shifted towards understanding the influence of Tumour Microenvironment (TME) in cancer prognosis. This study aims to uncover the relationship between TME, molecular profiles, and clinical outcomes in HGSC. To achieve this, Random Forest (RF) analysis on single-cell and spatially resolved data from high-plex immunofluorescent images was employed. By leveraging RF's feature selection capabilities, we identified pertinent TME features associated with the clinico-molecular characteristics of HGSC. Our analysis revealed distinctive TME characteristics in HGSC patients with BRCA loss and homologous recombination repair proficiency. Notably, we found that Major Histocompatibility Complex (MHC) class II expression, originating from cancer cells, was a critical feature that shaped the immune environment and cancer behaviour. Moreover, we discovered that MHC molecules, specifically MHC-II, were also crucial in distinguishing short and long-term survival groups. Increased MHC-II expression was associated with improved survival, independently of molecular profiles. This association highlighted the importance of endogenous immunity in fighting cancer. In conclusion, MHC-II is a crucial prognostic marker in our study, offering a valuable and assessable metric for patient prognosis. In addition, the identification of phenotypically distinct survival groups based on TME characteristics underscores the potential of our approach in enhancing patient stratification and guiding personalized treatment strategies, ultimately improving the management of HGSC.