Distinct Gene Signature for Early Differentiation of MASL and MASH via Machine Learning

Overview

This study identified a robust six-gene signature that accurately distinguishes metabolic dysfunction-associated simple steatosis (MASL) from steatohepatitis (MASH) using integrated transcriptomic data and machine learning. An artificial neural network model based on these genes demonstrated high diagnostic accuracy, validated across multiple cohorts.

Background

Metabolic dysfunction-associated steatotic liver disease (MASLD) encompasses a spectrum from simple steatosis (MASL) to steatohepatitis (MASH), with MASH representing a more severe, inflammatory state that can progress to fibrosis and liver failure. Current diagnostic methods, including liver biopsy and imaging, have limitations due to invasiveness, cost, and accessibility. Noninvasive molecular biomarkers capable of early and accurate differentiation between MASL and MASH are urgently needed to improve patient management and outcomes.

Data Highlights

The artificial neural network model based on these genes achieved an AUC of 0.893 (95% CI 0.854 to 0.925) in the validation cohort.

Key Findings

• Identification of 656 differentially expressed genes between MASL and MASH.
• Six key genes (MMP9, FABP5, TREM2, CTSD, UBD, MAP2K1) consistently selected by multiple machine learning algorithms.
• Five genes were upregulated in MASH, while MAP2K1 was downregulated.
• Individual genes showed moderate diagnostic performance with AUCs ranging from 0.692 to 0.822.
• The artificial neural network classifier based on these genes achieved high diagnostic accuracy with an AUC of 0.893 in an independent validation cohort.
• Immune infiltration analysis revealed increased monocytes, M0 and M1 macrophages, and activated dendritic cells in MASH, with key genes predominantly expressed in myeloid cell populations.

Clinical Implications

The identified six-gene signature offers a promising noninvasive molecular tool for early and accurate differentiation between MASL and MASH, potentially reducing reliance on invasive liver biopsies. Incorporation of this gene signature into diagnostic workflows could improve risk stratification and guide timely therapeutic interventions. Furthermore, the immune cell associations highlight potential targets for future therapeutic development.

Conclusion

This multicohort machine learning-based study establishes a distinct gene signature with high diagnostic accuracy for differentiating MASL from MASH and enhances understanding of immune-metabolic mechanisms driving disease progression.

References

1. Hasin Brumshtein et al. 2023 -- Multi-cohort transcriptomic integration identifies MASH-associated genes