Machine Learning-Based Integration of Single-Cell and Bulk Transcriptome Reveals Coagulation Signature and Phenotypic...

Area of Science:

• Oncology
• Genomics
• Computational Biology

Background:

• Hepatocellular carcinoma (HCC) is a leading cause of cancer death worldwide.
• The liver's role in coagulation is critical, yet single-cell mechanisms in HCC remain unclear.
• Existing prognostic models for HCC lack detailed single-cell insights into coagulation.

Purpose of the Study:

• To identify and characterize coagulation subtypes and heterogeneity in HCC cells at the single-cell level.
• To develop a novel, single-cell-based coagulation gene signature and risk score for HCC prognosis.
• To explore the underlying mechanisms of prognostic differences and potential therapeutic applications.

Main Methods:

• Utilized KEGG and GO databases for coagulation-related genes.
• Applied machine learning algorithms to define a single-cell coagulation gene signature.
• Constructed a Coagulation-Associated Risk Score (CARS) model in the TCGA-LIHC cohort.
• Integrated clinicopathological data and CARS for nomogram development.
• Performed pathway, cellular communication, and pseudotime trajectory analyses.

Main Results:

• Identified distinct coagulation subtypes and heterogeneity within HCC malignant cells.
• Developed a robust CARS model demonstrating accurate prognostic prediction in the TCGA-LIHC cohort.
• A nomogram integrating CARS and clinicopathological features provides individualized prognostic assessment.
• Dissected mechanisms driving prognostic variations linked to coagulation risk.

Conclusions:

• The developed CARS system offers precise prognostic prediction for HCC.
• This risk assessment provides a theoretical foundation for precision medicine approaches in HCC treatment.
• The study elucidates single-cell level coagulation mechanisms influencing HCC outcomes.