Development of a machine learning-based signature utilizing inflammatory response genes for predicting prognosis and immune microenvironment in ovarian cancer

Abstract

Abstract Ovarian cancer (OC) represents a significant health challenge, characterized by a particularly unfavorable prognosis for affected women. Accumulating evidence supports the notion that inflammation-related factors impacting the normal ovarian epithelium may contribute to the development of OC. However, the precise role of inflammatory response-related genes (IRRGs) in OC remains largely unknown. To address this gap, we performed an integration of mRNA expression profiles from 7 cohorts and conducted univariate Cox regression analysis to screen 26 IRRGs. By utilizing these IRRGs, we categorized patients into subtypes exhibiting diverse inflammatory responses, with subtype B displaying the most prominent immune infiltration. Notably, the elevated abundance of Treg cells within subtype B contributed to immune suppression, resulting in an unfavorable prognosis for these patients. Furthermore, we validated the distribution ratios of stromal cells, inflammatory cells, and tumor cells using whole-slide digitized histological slides. We also elucidated differences in the activation of biological pathways among subtypes. In addition, machine learning algorithms were employed to predict the likelihood of survival in OC patients based on the expression of prognostic IRRGs. Through rigorous testing of over 100 combinations, we identified CXCL10 as a crucial IRRG. Single-cell analysis and vitro experiments further confirmed the potential secretion of CXCL10 by macrophages and its involvement in lymphangiogenesis within the tumor microenvironment. Overall, the study provides new insights into the role of IRRGs in OC and may have important implications for the development of novel therapeutic approaches.