netSmooth

netSmooth applies network-diffusion smoothing to single-cell RNA-seq (scRNA-seq) gene expression using biological network priors such as protein-protein interaction networks to improve signal recovery for downstream analyses like clustering and differential expression.

Key Features:

• Network-diffusion approach: Uses biological networks, e.g., protein-protein interactions, as priors on gene co-expression and applies network diffusion that accounts for the covariance structure to smooth expression values.
• Improved clustering and cell-type resolution: Enhances clustering of scRNA-seq data and facilitates distinguishing distinct cell populations from noisy, sparse data to aid cell type identification.
• Applicability across experimental designs: Demonstrated improvements in analyses involving distinct cell populations, time-course experiments, and cancer genomics.
• Dropout mitigation and downstream readiness: Mitigates scRNA-seq dropouts by leveraging known gene relationships and produces smoothed data suitable for downstream analyses such as differential expression.

Scientific Applications:

• Cell type identification: Refines scRNA-seq data to improve precise cell type identification.
• Developmental biology: Resolves cellular heterogeneity in developmental biology studies.
• Oncology / Cancer genomics: Distinguishes cancerous versus non-cancerous cells in cancer genomics and oncology research.
• Time-course experiments: Improves temporal profiling in time-course scRNA-seq experiments.

Methodology:

Performs network-diffusion smoothing of gene expression by incorporating biological networks (e.g., protein-protein interactions) as priors and leveraging the covariance structure to reduce dropouts and improve data for clustering and differential expression analyses.