FEAST

FEAST selects representative gene features to improve cell clustering accuracy in single-cell RNA sequencing (scRNA-seq) data analysis.

Key Features:

• Feature Selection: Identifies subsets of genes whose expression patterns distinguish different cell types for downstream clustering.
• Plug-in Compatibility: Integrates with established clustering algorithms including SC3, TSCAN, SHARP, SIMLR, and Seurat.
• Consensus Clustering: Employs consensus clustering as the first step of its core algorithm.
• Gene-level Significance Inference: Performs statistical inference at the gene level as the second step of its core algorithm.
• Feature Set Validation: Validates an optimized feature set as the third step of its core algorithm.

Scientific Applications:

• scRNA-seq Cell Clustering: Enhances clustering accuracy by selecting high-quality feature genes prior to the core clustering step.

Methodology:

The core FEAST algorithm consists of three steps: consensus clustering, gene-level significance inference, and validation of an optimized feature set.