Work

Multi-rank non-negative matrix factorization reveals myeloid-centered co-expression remodeling and prioritizes MMP9 as a candidate therapeutic target in kawasaki disease

Year: 2026
Type: article
Abstract: Kawasaki disease (KD) causes coronary artery aneurysms in up to 25% of untreated children, but its transcriptomic resemblance to common febrile illnesses complicates early diagnosis. Current treatments are not directed at a defined disease mechanism, and 10–20% of patients do not respond to initial intravenous immunoglobulin. Using three GEO whole-blood cohorts covering KD, bacterial and viral infection, Henoch-Schönlein purpura, and juvenile idiopathic arthritis, we combined multi-rank non-negative matrix factorization with differential gene correlation analysis to identify hub genes whose co-expression with partner genes differs in KD. KD shared far more differentially expressed genes with bacterial infection (Jaccard index 0.504) than with viral (0.155) or autoimmune (0.017) mimics. Five MCC-ranked hub genes (ITGAM, MMP9, HK3, FCER1G, and CSF3R) were upregulated in KD and showed significant co-expression pairs in KD that were absent in mimics; immune deconvolution hub-gene expression correlated positively with estimated neutrophil fractions and negatively correlated with estimated CD8 + T-cell fractions (HK3: Spearman r = − 0.81). A 5-gene classifier built with TabPFN-V2.5 reached an AUC of 0.889 in external validation. Molecular docking gave S-3304 the most favourable predicted binding energy for MMP9 (− 10.21 kcal/mol), and can used as a candidate drug for the treatment of coronary artery injury and be used in biochemical and preclinical evaluations.
Language: English
Field: Medicine
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7226024319
FWCI:
Cites: 1
Cited by:
Related to: 10