SiCmiR Atlas: Single-Cell miRNA Landscape Reveals Hub-miRNA and Network Signatures in Human Cancers

Area of Science:

• Genomics and Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• MicroRNAs (miRNAs) are key regulators of gene expression.
• Single-cell miRNA profiling is technically challenging, limiting biological insights.
• Existing methods struggle with data sparsity and dropout in single-cell RNA sequencing (scRNA-seq).

Purpose of the Study:

• To develop a computational method for predicting miRNA expression at the single-cell level.
• To overcome technical limitations in small RNA sequencing.
• To create a comprehensive resource for single-cell miRNA analysis and biomarker discovery.

Main Methods:

• Developed SiCmiR, a two-layer neural network using 977 landmark genes to predict miRNA expression.
• Trained SiCmiR on 6,462 TCGA paired miRNA-mRNA samples.
• Constructed SiCmiR-Atlas, a database of 362 public datasets (9.36 million cells, 726 cell types).

Main Results:

• SiCmiR accurately predicts miRNA expression profiles from limited gene data, reducing scRNA-seq dropout sensitivity.
• Identified candidate hub miRNAs in various cancers (hepatocellular carcinoma, pancreatic ductal carcinoma) and extracellular vesicle-mediated crosstalk in glioblastoma.
• SiCmiR demonstrates state-of-the-art accuracy and generalizability across cancer types and drug perturbations.
• SiCmiR-Atlas provides interactive visualization, biomarker identification, and miRNA-target networks.

Conclusions:

• SiCmiR effectively translates bulk-derived statistical power to a single-cell view of miRNA biology.
• SiCmiR-Atlas serves as a valuable community resource for single-cell miRNA research and biomarker discovery.
• The developed method overcomes previous technical barriers, enabling deeper understanding of miRNA functions in cellular processes and diseases.