A machine learning-derived gene signature for assessing rupture risk and circulatory immunopathologic landscape in patients with intracranial aneurysms

Abstract

BackgroundIntracranial aneurysm (IA) is an uncommon but severe subtype of cerebrovascular disease, with high mortality after aneurysm rupture. Current risk assessments are mainly based on clinical and imaging data. This study aimed to develop a molecular assay tool for optimizing the IA risk monitoring system.MethodsPeripheral blood gene expression datasets obtained from the Gene Expression Omnibus were integrated into a discovery cohort. Weighted gene co-expression network analysis (WGCNA) and machine learning integrative approaches were utilized to construct a risk signature. QRT-PCR assay was performed to validate the model in an in-house cohort. Immunopathological features were estimated using bioinformatics methods.ResultsA four-gene machine learning-derived gene signature (MLDGS) was constructed for identifying patients with IA rupture. The AUC of MLDGS was 1.00 and 0.88 in discovery and validation cohorts, respectively. Calibration curve and decision curve analysis also confirmed the good performance of the MLDGS model. MLDGS was remarkably correlated with the circulating immunopathologic landscape. Higher MLDGS scores may represent higher abundance of innate immune cells, lower abundance of adaptive immune cells, and worse vascular stability.ConclusionsThe MLDGS provides a promising molecular assay panel for identifying patients with adverse immunopathological features and high risk of aneurysm rupture, contributing to advances in IA precision medicine.