# Integrative molecular subtyping and diagnostic model construction for Moyamoya disease based on diverse programmed cell death gene patterns

## TL;DR

This study explores how programmed cell death and immune responses contribute to Moyamoya disease, identifying molecular subtypes and potential diagnostic markers.

## Contribution

The study introduces a novel integrative approach combining gene expression, immune infiltration, and machine learning to subtype MMD and identify diagnostic biomarkers.

## Key findings

- Two distinct molecular subgroups were identified in MMD patients through unsupervised clustering.
- Key genes like ADRB2, IL1B, and CASP4 were upregulated in one subgroup, linked to apoptosis and immune processes.
- An SVM-based diagnostic model achieved perfect AUC, with PCR validation confirming upregulation of model genes in MMD.

## Abstract

The role of cell death in the pathogenesis of Moyamoya disease (MMD) remains unclear.

Gene expression data from two publicly available MMD datasets (GSE157628 and GSE189993) were integrated. Differential expression analysis of cell death-related genes was performed, followed by immune cell infiltration analysis using the CIBERSORT algorithm. Unsupervised clustering identified molecular subgroups within MMD patients, and these were further analyzed using Gene Set Variation Analysis (GSVA) and Weighted Gene Co-expression Network Analysis (WGCNA). Diagnostic models were constructed using machine learning algorithms, and key model genes were validated by qRT-PCR in arterial samples from MMD patients.

Significant differences in gene expression were observed, with immune cell infiltration analysis showing differences in T follicular helper cells, activated dendritic cells, resting and activated mast cells, and eosinophils. Unsupervised clustering identified two distinct patient groups, and differential gene expression revealed upregulation of genes like ADRB2, FGR, ICAM1, IL1B, DDIT4, CXCL1, CASP4, G0S2, CYP1B1, and CD74 in C2 cluster. Weighted gene co-expression network analysis (WGCNA) identified key genes in the most correlated module. Machine learning models (RF, SVM, GLM, XGB) were constructed, with the SVM model showing the best performance (AUC = 1.000). A nomogram based on the SVM model demonstrated high predictive accuracy. Enrichment analysis revealed that key genes were involved in apoptosis and antigen processing, with strong diagnostic performance confirmed by ROC curves. PCR analysis revealed that model genes RGS1, MUC1, KCNA2, TAC1, and SOST were all up-regulated in MMD group.

This study provides novel insights into the molecular landscape of MMD, highlighting the importance of cell death-related pathways and immune responses. The identified biomarkers and molecular subgroups offer potential targets for therapeutic intervention and improved diagnostic strategies in MMD.

The online version contains supplementary material available at 10.1186/s13023-025-03816-y.