Alzheimer's disease-associated transcriptional signatures define prognostic subtypes in glioma

Gliomas are molecularly heterogeneous central nervous system tumors with marked variation in clinical outcome. Alzheimer's disease (AD)-associated transcriptional alterations may capture neural and immune programs relevant to glioma biology, but they do not by themselves establish a direct mechanistic relationship between AD and glioma.

AD-associated differentially expressed genes were identified from GSE132903 and evaluated in TCGA/GTEx and CGGA glioma datasets. Consensus clustering, enrichment analysis, Cox and LASSO modeling, immune deconvolution, and mutation analyses were performed. The incremental prognostic value of the AD-Driven score (ADDs) was tested after adjustment for age, sex, WHO grade, IDH status, 1p/19q codeletion, and MGMT promoter methylation. Protein-expression and glioma survival annotations for the 14 model genes were reviewed in the Human Protein Atlas (HPA).

We identified 470 CE-associated differentially expressed genes enriched in synaptic, calcium-signaling, and neurotransmitter pathways. A 14-gene ADDs model stratified survival in the discovery and validation cohorts. Among 554 complete cases with 147 deaths, continuous standardized ADDs remained associated with overall survival after molecular and clinicopathological adjustment (HR per SD = 1.76, 95 % CI 1.26-2.44, P = 0.0008), improving model fit (likelihood-ratio P = 0.0008) but only modestly increasing the C-index (0.876 to 0.880). HPA review provided directionally supportive glioma survival annotations for seven model genes.

The ADDs captures prognostically relevant neural and immune transcriptional variation in glioma and provides incremental information beyond established molecular variables when modeled continuously. The cross-cancer IMvigor210 analysis remains exploratory and does not establish prediction of immunotherapy benefit in glioma. HPA observations provide orthogonal public-database support for selected signature components, while independent tissue-based, mechanistic, and glioma-specific treatment-response validation remain necessary.