PRIME

PRIME implements probabilistic imputation to reduce dropout effects in single-cell RNA sequencing (scRNA-seq) expression profiles, improving transcriptome estimates for downstream analyses.

Key Features:

• Probabilistic imputation: Applies a probabilistic framework to estimate missing or zero-inflated expression values caused by dropouts in scRNA-seq data.
• Cell correspondence network: Constructs a cell correspondence network that identifies and groups cells based on their transcriptome profiles.
• Local subnetwork aggregation: Focuses on local subnetworks of similar cell types to aggregate information from neighboring cells.
• Expression adjustment: Adjusts gene expression estimates within localized networks by leveraging collective information from neighboring cells to infer more accurate expression levels.
• Dropout reduction: Mitigates dropout effects and reduces zero inflation to recover biological signals obscured by technical noise.
• Benchmarking: Validated using synthetic datasets and eight real single-cell sequencing datasets.
• Downstream improvement: Improves data visualization, increases clustering accuracy, and enables discovery of previously obscured gene expression patterns.

Scientific Applications:

• Cell type identification: Enables identification of novel cell types by recovering expression signals obscured by dropouts.
• Trajectory and differentiation analysis: Supports analysis of cellular differentiation pathways and lineage inference through improved expression estimates.
• Disease mechanism exploration: Facilitates exploration of disease mechanisms at single-cell resolution by enhancing signal detection.
• Visualization and clustering: Enhances data visualization and clustering to aid interpretation of cellular heterogeneity.

Methodology:

Constructs a cell correspondence network, identifies and groups cells by transcriptome similarity, focuses on local subnetworks of similar cell types, and performs probabilistic imputation by adjusting gene expression estimates using information from neighboring cells.