A PIK3CA and NF1 expression-based prognostic signature derived from Mendelian randomization identifies causal immune-regulatory determinants of glioblastoma outcomes

The causal roles of PIK3CA and NF1 expression in glioblastoma multiforme (GBM) pathogenesis and their immune-regulatory functions remain poorly understood. We developed and validated an expression-based prognostic signature to predict survival and assess immunotherapy-related biomarker potential in GBM patients. We employed two-sample Mendelian randomization (MR) analysis using genome-wide association study data to establish causal relationships between genetically predicted PIK3CA and NF1 expression levels and GBM risk. Whole-exome sequencing was performed on GBM tumor samples to identify differentially expressed genes (DEGs, FDR < 0.05) associated with PIK3CA and NF1 alterations. A 6-gene expression-based prognostic signature comprising PIK3CA, NF1, PGPEP1, COL5A1, MFGE8, and MSH6 was constructed using multivariate Cox regression in The Cancer Genome Atlas (TCGA) GBM cohort and validated across independent Gene Expression Omnibus (GEO) datasets. As proof-of-concept for the immune-regulatory basis of the signature, we evaluated its trans-tumor applicability in predicting immunotherapy response using the IMvigor210 cohort. MR analysis demonstrated that genetically predicted higher PIK3CA expression in normal tissues was causally associated with decreased GBM risk (IVW OR = 0.176, 95% CI 0.056–0.556, P = 0.003), while NF1 showed a suggestive protective trend. Notably, higher NF1 expression also correlated with increased ankylosing spondylitis risk (OR = 0.999, P = 0.042), revealing pleiotropic immunomodulatory functions that validate its immune-regulatory role independent of somatic tumor mutations. The 6-gene expression signature effectively stratified patients into high-risk and low-risk groups with significant survival differences across all validation cohorts (TCGA: log-rank P = 0.037; GEO cohorts: P = 0.012–0.041). The signature demonstrated robust discriminative ability (C-index: 0.71 in TCGA training set, 0.64–0.66 in GEO validation sets) and time-dependent predictive performance, significantly outperforming clinical variables including age (AUC = 0.668), gender (AUC = 0.321), and tumor grade (AUC = 0.500). In the IMvigor210 cohort, the expression signature showed trans-tumor applicability for immune checkpoint blockade response prediction (ROC AUC = 0.566, 95% CI 0.487–0.645), with low-risk patients demonstrating superior progression-free survival compared to high-risk patients (P = 0.016), providing proof-of-concept evidence for the immune-regulatory biological basis of the signature. Our integrated MR and multi-omics approach establishes PIK3CA and NF1 at the germline expression level as causal determinants of GBM risk at the germline expression level through immune-regulatory mechanisms, distinct from their somatic mutation effects in established tumors. We identified a clinically robust 6-gene expression signature that integrates complementary biological dimensions—oncogenic pathways, extracellular matrix remodeling, immune regulation, and genomic instability—for prognostic stratification in GBM. While the trans-tumor immunotherapy prediction requires GBM-specific validation, our findings provide a foundation for precision medicine approaches and hypothesis-driven biomarker development in GBM.