findPC

findPC selects the optimal number of principal components for single-cell RNA-seq data to improve the accuracy and reproducibility of downstream analyses such as clustering, differential expression, and trajectory inference.

Key Features:

• Automated Selection: Automates selection of the optimal number of principal components by replacing subjective visual inspection of elbow plots.
• Multiple Algorithms: Implements six distinct algorithms that refine the elbow method by automatically identifying the elbow (inflection) point in a scree plot.
• Performance Evaluation: Validated on real single-cell RNA-seq datasets from multiple human and mouse tissues and cell types, with the perpendicular line method at 30 PCs reported as having superior overall performance compared to manual selection.

Scientific Applications:

• Cell Type Identification: Provides objective dimensionality reduction inputs that improve clustering-based identification of cell types in scRNA-seq data.
• Gene Expression Analysis: Supports differential gene expression analyses by supplying consistent principal component selection for downstream statistical testing.
• Trajectory Inference: Aids reconstruction of lineage trajectories and cellular differentiation pathways by determining appropriate PCs for pseudotime and trajectory methods.

Methodology:

Six computational methods refine the traditional elbow method by evaluating the inflection point in a scree plot, including a perpendicular line method evaluated at 30 PCs.