cSVI-subtype: a GWAS-anchored, vascular-centered framework for molecular subtype barcode discovery in cerebral small vessel disease

Motivation: Cerebral small vessel disease (cSVD) is still classified mainly by clinical phenotypes and neuroimaging features, which provide limited molecular resolution for robust and transferable subtype definition. We developed cSVI-subtype to derive compact and reproducible molecular subtype barcodes for cSVD by integrating human genetics with cross-species single-cell evidence. Results: cSVI-subtype is a GWAS-anchored, vascular-centered framework for molecular subtype barcode discovery. It prioritizes human candidate genes through colocalization and supportive Mendelian randomization, and then projects these signals into cross-species single-cell disease models. The framework integrates a layered vascular evidence chain comprising vascular abundance gating (CSS), source-to-vascular interaction strength (CIS), and ligand-receptor-to-gene allocation with pathway/GO projection (LRS), enabling subtype-specific prioritization while mitigating receptor-level sparsity. Applied to cSVD, cSVI-subtype identified compact and reproducible barcodes for A1/N3KO-like and B1/BCAS-like states and remained stable across GWAS perturbation, single-chain ablation, and internal control analyses. These results support cSVI-subtype as a practical framework for molecular subtype definition in cSVD and suggest its potential extensibility to other vascular-related disorders. Availability and implementation: cSVI-subtype is implemented in R and distributed as the package csviSubtype (version 0.1.0). Source code, documentation, shipped example input files, and a quickstart vignette are freely available at https://github.com/YuqianLii/csviSubtype.