scFeatureFilter

scFeatureFilter selects informative genes from single-cell RNA-seq datasets by using a correlation-based framework to remove genes dominated by technical variability and mitigate high drop-out rates and technical noise.

Key Features:

• Correlation-based filtering: Quantifies gene–gene correlation across the transcriptome and interprets low-correlation features as dominated by technical variability rather than biological signal.
• Drop-out and technical noise mitigation: Targets features affected by high drop-out rates and substantial technical noise for removal from the expression matrix.
• Filtered expression matrix: Produces a filtered expression matrix enriched for genes whose variation reflects cellular heterogeneity rather than technical artifacts.
• Improves downstream analyses: Reduces systematic feature-level noise to improve outcomes for clustering, trajectory inference, and differential expression analyses.
• R package implementation: Provided as an R package to integrate as a preprocessing step in standard scRNA-seq analysis workflows.

Scientific Applications:

• Clustering: Preprocesses gene features to improve identification of cell populations via clustering methods.
• Trajectory inference: Enhances trajectory and pseudotime analyses by retaining genes with coherent transcriptomic correlation structure.
• Differential expression: Improves interpretability of differential expression results by filtering out genes dominated by technical variability.

Methodology:

Computes gene–gene correlations across the transcriptome, quantifies lack of correlation per gene, and removes genes with little to no correlation to yield a filtered expression matrix.