Related Experiment Video
Updated: Aug 15, 2026
Biobanking of Human Aqueous and Vitreous Liquid Biopsies for Molecular Analyses
Published on: September 11, 2023
Aqueous Humor Proteomics Reveals the Molecular Basis for Differential Treatment Responses in nAMD and pmCNV
Hanqing Zhao1,2,3, Benbo Gao4, Ning Ding1,2,3
1Beijing Tongren Eye Center, Beijing Key Laboratory of Intraocular Tumor Diagnosis and Treatment, Beijing Tongren Hospital, Capital Medical University, Beijing, China.
Purpose:
Choroidal neovascularization (CNV) causes vision loss in neovascular age-related macular degeneration (nAMD) and pathologic myopia-related CNV (pmCNV). Despite shared clinical features, they respond differently to treatment. This study compared aqueous humor (AH) proteomic profiles of nAMD and pmCNV to identify biomarkers and mechanisms of CNV formation.
Methods:
AH samples were collected from eyes with nAMD, pmCNV, age-related cataract controls, and pathologic myopia without CNV. Label-free data-independent acquisition (DIA) mass spectrometry was used for proteomic profiling. Differentially expressed proteins were identified using limma models adjusted for age, sex, and axial length. Functional enrichment analyses used Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Reactome databases. Stability-based feature selection, LightGBM modeling, and SHapley Additive exPlanations (SHAP) prioritized discriminative proteins; selected findings were evaluated in an independent nAMD-control Olink cohort.
Results:
DIA proteomics identified 3986 proteins, of which 3667 passed quality filtering. nAMD and pmCNV showed weakly correlated proteomic alterations (r2 = 0.061), with 385 nAMD-specific, 398 pmCNV-specific, and 100 shared altered proteins. nAMD was characterized by vascular endothelial growth factor (VEGF/VEGF receptor [VEGFR]) pathway activation and reduced glucose-metabolism-related proteins, whereas pmCNV showed extracellular matrix remodeling, platelet/coagulation activation, and downregulation of Slit/Robo-related neurovascular guidance pathways. The stability-selected LightGBM model using 33 proteins achieved a macro-average area under the curve (AUC) of 0.92. SHAP analysis highlighted VEGFR2 (KDR) and VEGFA as major nAMD-associated drivers, and Olink validation showed concordant expression trends for six selected proteins.
Conclusions:
nAMD and pmCNV exhibit largely distinct AH proteomic signatures, supporting different molecular mechanisms of CNV formation and anti-VEGF responsiveness. AH proteomics may help prioritize subtype-specific candidate biomarkers, warranting further validation across independent cohorts.
