A reproducible dual-axis chemokine co-expression architecture in glioblastoma across three independent transcriptomic cohorts, dissociated from non-reproducible, IDH-confounded survival associations

Opposing chemokine programs shape the glioblastoma (GBM) microenvironment, but it is unclear which features are reproducible once mathematical circularity is removed, and how strongly IDH status confounds chemokine–survival associations. We analysed 25 chemokines/receptors in three GBM cohorts—TCGA ( n = 159), CGGA ( n = 85), GSE16011 ( n = 159)—across RNA-seq and microarray. Co-expression was assessed by leave-one-gene-out (LOGO) correlation and, independently, by principal component analysis, factor analysis and network community detection, each performed separately in every cohort; reproducibility by Fisher r-to-z random-effects meta-analysis with REML and Hartung–Knapp sensitivity analyses (k = 3). Infiltration used independent, non-chemokine marker signatures. Survival analysis was restricted to IDH-wild-type GBM (per-1-SD Cox, FDR-corrected). The dual-axis architecture—a CXCR3-ligand module (CXCL9/CXCL10/CXCL11) opposed by an ELR⁺ CXCL/CCL20 myeloid module—survived LOGO correction and was recovered as the second principal component and as distinct network communities in each cohort independently (PC2 loading agreement ρ = 0.87–0.89). The first component was a shared infiltration gradient loading both modules positively, so they are not mutually exclusive. The modules were not equally reproducible: the CXCR3-ligand module showed minimal heterogeneity and excluded the null even under Hartung–Knapp (CXCL11 r = 0.56, I²=0%), whereas the ELR⁺/myeloid genes were directionally consistent but heterogeneous in magnitude (CCL20 r = − 0.43, I²=84%). Survival associations were fragile: CXCL10/CXCL11 were nominally protective (HR 0.70/0.73) but failed FDR (q = 0.19), with opposite directions in TCGA and CGGA (I²=91%). The co-expression architecture is reproducible—quantitatively for the CXCR3-ligand module, directionally for the ELR⁺/myeloid module—whereas chemokine–survival associations are non-reproducible, IDH-confounded and FDR-null.

Authors

Institutions

Publication Details

Journal
Discover Oncology
Published
2026-11-16
DOI
https://doi.org/10.1007/s12672-026-06131-5
Primary Topic
Glioma Diagnosis and Treatment
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A reproducible dual-axis chemokine co-expression architecture in glioblastoma across three independent transcriptomic cohorts, dissociated from non-reproducible, IDH-confounded survival associations

Peng Chan Huang, Man Chong U
3 links
Discover Oncology
Glioma Diagnosis and Treatment
article

A reproducible dual-axis chemokine co-expression architecture in glioblastoma across three independent transcriptomic cohorts, dissociated from non-reproducible, IDH-confounded survival associations

Peng Chan Huang, Man Chong U
article en

Abstract

Opposing chemokine programs shape the glioblastoma (GBM) microenvironment, but it is unclear which features are reproducible once mathematical circularity is removed, and how strongly IDH status confounds chemokine–survival associations. We analysed 25 chemokines/receptors in three GBM cohorts—TCGA ( n = 159), CGGA ( n = 85), GSE16011 ( n = 159)—across RNA-seq and microarray. Co-expression was assessed by leave-one-gene-out (LOGO) correlation and, independently, by principal component analysis, factor analysis and network community detection, each performed separately in every cohort; reproducibility by Fisher r-to-z random-effects meta-analysis with REML and Hartung–Knapp sensitivity analyses (k = 3). Infiltration used independent, non-chemokine marker signatures. Survival analysis was restricted to IDH-wild-type GBM (per-1-SD Cox, FDR-corrected). The dual-axis architecture—a CXCR3-ligand module (CXCL9/CXCL10/CXCL11) opposed by an ELR⁺ CXCL/CCL20 myeloid module—survived LOGO correction and was recovered as the second principal component and as distinct network communities in each cohort independently (PC2 loading agreement ρ = 0.87–0.89). The first component was a shared infiltration gradient loading both modules positively, so they are not mutually exclusive. The modules were not equally reproducible: the CXCR3-ligand module showed minimal heterogeneity and excluded the null even under Hartung–Knapp (CXCL11 r = 0.56, I²=0%), whereas the ELR⁺/myeloid genes were directionally consistent but heterogeneous in magnitude (CCL20 r = − 0.43, I²=84%). Survival associations were fragile: CXCL10/CXCL11 were nominally protective (HR 0.70/0.73) but failed FDR (q = 0.19), with opposite directions in TCGA and CGGA (I²=91%). The co-expression architecture is reproducible—quantitatively for the CXCR3-ligand module, directionally for the ELR⁺/myeloid module—whereas chemokine–survival associations are non-reproducible, IDH-confounded and FDR-null.

Discover Oncology
Mackay Memorial Hospital (TW)
Good health and well-being
Glioma Diagnosis and Treatment
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon