VAM

VAM computes cell-level gene set (pathway) scores for single-cell RNA sequencing (scRNA-seq) data to enable pathway-level analysis and statistical inference that accounts for technical noise and data sparsity.

Key Features:

• Integration with Seurat: Implements outputs compatible with the Seurat framework for incorporation into single-cell analysis workflows.
• Cell-specific pathway scoring: Transforms a cell-by-gene matrix into a cell-by-pathway matrix by computing pathway scores for individual cells.
• Handling technical noise and sparsity: Adjusts for high technical noise, inflated zero counts, sparsity, and large sample sizes characteristic of scRNA-seq data.
• Gamma approximation for null distribution: Uses an accurate gamma approximation for the distribution of pathway scores under the null hypothesis of uncorrelated technical noise to support population- and cell-level inference.
• Demonstrated performance: Shown in simulations and real scRNA-seq datasets to provide superior classification accuracy and reduced computational cost compared to existing single-sample gene set testing approaches.

Scientific Applications:

• Cellular heterogeneity analysis: Provides cell-level pathway scores to resolve heterogeneity in scRNA-seq datasets.
• Developmental biology: Enables analysis of pathway dynamics across differentiating cell populations using scRNA-seq.
• Cancer research: Supports assessment of pathway activity variation at single-cell resolution in cancer scRNA-seq studies.
• Immunology: Profiles pathway activation states in immune cell populations measured by scRNA-seq.

Methodology:

Computes variance-adjusted Mahalanobis distances (modified Mahalanobis), derives pathway scores from the distribution of squared modified Mahalanobis distances, applies a gamma approximation for the null distribution, and provides density estimates under simulated sparsity conditions (e.g., 0.5 and 0.8).