Modelling the liver's regenerative capacity across different clinical conditions.

Area of Science:

• Hepatology
• Systems Biology
• Machine Learning

Background:

• Liver regeneration is crucial for recovery but can be hindered by age, metabolic disorders, fibrosis, and immunosuppression.
• Identifying biomarkers for regeneration under diverse conditions is essential for clinical application.

Purpose of the Study:

• To identify key transcriptomic, proteomic, and serum biomarkers of liver regeneration.
• To develop a predictive framework for regenerative capacity using systems biology and machine learning.

Main Methods:

• Utilized six mouse models representing diverse clinical contexts (age, fibrosis, steatosis, immunosuppression).
• Employed a novel contrastive deep learning framework with triplet loss to analyze regenerative trajectories.
• Integrated transcriptomic, proteomic, and clinical data for predictive modeling.

Main Results:

• Regeneration was significantly delayed in aged, steatotic, and fibrotic models, evidenced by reduced Ki-67 staining.
• Transcriptomic and proteomic analyses consistently showed downregulation of cell cycle genes in impaired regeneration.
• A deep learning model accurately predicted regenerative outcomes (87.9%), highlighting six key genes (Wee1, Rbl1, Gnl3, Mdm2, Cdk2, Ccne2).

Conclusions:

• Identified conserved cell cycle regulators critical for efficient liver regeneration.
• Developed a predictive framework for assessing regenerative capacity, applicable across species.
• This multi-omics and deep learning approach offers a promising strategy to understand and potentially enhance liver regeneration in complex clinical settings.