scBPGRN: Integrating single-cell multi-omics data to construct gene regulatory networks based on BP neural network

Area of Science:

• Genomics
• Bioinformatics
• Cancer Biology

Background:

• Cancer progression involves complex genomic and epigenetic changes.
• Integrating multi-omics data is crucial for understanding cancer pathogenesis.
• Single-cell resolution offers deeper insights into cellular heterogeneity.

Purpose of the Study:

• To develop a novel algorithm, scBPGRN, for constructing gene regulatory networks using single-cell RNA sequencing and DNA methylation data.
• To identify key genes associated with hepatocellular carcinoma (HCC) by analyzing gene regulatory networks.
• To provide a framework for integrating single-cell multi-omics data for cancer research.

Main Methods:

• Developed the single-cell back-propagation gene regulatory network (scBPGRN) algorithm.
• Utilized biweight extreme correlation coefficients for factor correlation measurement.
• Employed neural networks to calculate generalized weights for network construction.
• Calculated node strength to identify cancer-associated genes.

Main Results:

• Applied scBPGRN to HCC data, constructing a regulatory network and identifying key genes (e.g., MYCBP, KLHL35, PRKCZ, SERPINA6).
• HCC-related genes were concentrated within the top-ranked genes.
• Analysis of identified subpopulations revealed concentrated HCC-related genes.
• Functional enrichment analysis validated the constructed gene regulatory network.

Conclusions:

• The scBPGRN algorithm effectively integrates single-cell multi-omics data to construct gene regulatory networks.
• Identified key genes provide potential targets for understanding and treating HCC.
• The study offers a valuable reference for future single-cell multi-omics integration in cancer research.