Multi-omics profiling uncovers diagnostic biomarker panel, progression-associated subtypes, and prognostic signature in pediatric mitochondrial disease

Abstract

Pediatric mitochondrial disease (PMD) is challenging due to its genetic heterogeneity, diagnostic difficulty, and unpredictable trajectories. Current biomarkers lack sensitivity to diagnose PMD or predict its severe course. Here, we show a multi-omics framework with longitudinal monitoring to resolve these challenges. We identify metabolic and transcriptomic dysregulation in PMD, deriving a robust 7-feature multi-omics diagnostic panel (6 metabolites, 1 transcript). In an independent validation cohort, this panel achieved superior diagnostic accuracy (AUC = 0.96) compared to conventional biomarkers. Beyond diagnosis, we addressed the need for patient stratification. Unsupervised proteomic clustering revealed two subtypes (Cluster I and II) that diverged over time; Cluster II patients had downregulation of cytoskeletal and immune pathways and worse clinical outcomes at 12 months. We leveraged these insights to develop a 12-transcript prognostic signature that predicts clinical prognostication with high accuracy (AUC = 0.85). This study establishes a molecular framework for PMD, offering validated tools for precision diagnosis and prognostic stratification. Pediatric mitochondrial disease lacks reliable biomarkers. Here, the authors show that a multi-omics study identified a 7-feature diagnostic panel (AUC = 0.96), revealed two proteomic subtypes with differing outcomes, and developed a 12-transcript prognostic signature (AUC = 0.85).

The authors' abstract, as published at the source. Nature Communications, 2026 · DOI ↗