Machine learning algorithms for predicting glioma patient prognosis based on CD163+FPR3+ macrophage signature

Abstract

Tumor-associated macrophages (TAMs) play a vital role in glioma progression and are associated with poor outcomes in glioma patients. However, the specific roles of different subpopulations of TAMs remain poorly understood. Two distinct cell types, glioma and myeloid cells, were identified through single-cell sequencing analysis in gliomas. Within the TAMs-associated weighted gene co-expression network analysis (WGCNA) module, FPR3 emerged as a hub gene and was found to be expressed on CD163+ macrophages, while also being associated with clinical outcomes. Subsequently, a comprehensive assessment was undertaken to investigate the correlation between FPR3 expression and immune characteristics, revealing that FPR3 potentially plays a role in reshaping the glioma microenvironment. We identified a macrophage subset with the nonzero expression of CD163 and FPR3 (CD163+FPR3+). Using the expression profiles of CD163+FPR3+ macrophage-related signature, we employed ten machine learning algorithms to construct a prognostic model across six glioma cohorts. Subsequently, we employed an optimal algorithm to generate an artificial intelligence-driven prognostic signature specifically for CD163+FPR3+ macrophages. The development of this model was based on the average C-index observed in the aforementioned six cohorts. The risk score of this model consistently and effectively predicted overall survival, surpassing the accuracy of conventional clinical factors and 100 previously published signatures. Consequently, the CD163+FPR3+ macrophage-related score shows potential as a prognostic biomarker for glioma patients.