Elucidating the Role of Pyroptosis in Lower-Grade Glioma: Development of a Novel Scoring System to Enhance Personalized Therapeutic Approaches

Abstract

Pyroptosis, an orchestrated cellular death pathway, has gained attention due to its role in the pathophysiology and evolution of numerous malignancies. Despite this, no robust quantitative measure of pyroptosis activity in lower-grade glioma (LGG) exists currently. We scrutinized the transcriptomic data of LGG specimens acquired from TCGA and CGGA repositories, juxtaposed with the expression patterns of healthy brain tissues from the GTEx database. A register of pyroptosis-associated genes was extracted from the GSEA database. Utilizing unsupervised clustering algorithms on the expression patterns of these genes, we stratified LGG samples into unique subgroups. We implemented the Boruta machine learning algorithm to discern representative variables for each pyroptosis subtype and applied principal component analysis (PCA) to condense the dimensionality of the feature gene expression data, which led to the formulation of a pyroptosis scoring system (P score) to estimate pyroptosis activity in LGG. Furthermore, we affirmed the capacity of the P score to discriminate diverse cell subpopulations within a single-cell database and explored the correlations between the P score and clinical attributes, prognostic implications, and the tumor immune microenvironment in LGG. We identified three distinctive pyroptosis patterns with significant correlations to patient survival, clinicopathological properties, and characteristics of the tumor immune microenvironment (TIME). Two gene clusters, associated with unique prognostic and TIME attributes, emerged from differentially expressed genes (DEGs) across the pyroptosis patterns. The P score was formulated and authenticated as an autonomous prognostic determinant for overall survival in the TCGA and CGGA cohorts. Additionally, the P score demonstrated its competency to quantitatively represent pyroptosis activity across different cellular subpopulations in single-cell data. Notably, the P score in LGG was found to be indicative of tumor stemness and could serve as a predictive biomarker for the efficacy of temozolomide treatment and immunotherapy, underscoring its potential clinical utility. Our investigation pioneers a novel pyroptosis-centric scoring system with significant prognostic implications. The P score holds promise as a potential predictive biomarker for the response to chemotherapy and immunotherapy, facilitating the development of personalized therapeutic approaches in LGG patients.

Keywords: Lower-grade glioma; Machine learning; Pyroptosis; Single-cell RNA-seq; Tumor microenvironment.