scclusteval

scclusteval evaluates cluster stability in single-cell RNA sequencing (scRNAseq) datasets to quantify robustness of unsupervised clustering and guide selection of clustering parameters.

Key Features:

• Cluster Stability Evaluation: Uses a subsampling approach that repeatedly clusters subsets of cells to assess consistency of clusters across runs and parameter settings.
• Jaccard Similarity Index: Quantifies cluster stability by comparing clustering solutions using the Jaccard similarity index.
• Snakemake Workflow Integration: Integrates with a Snakemake workflow that automates subsampling and repeated clustering and enables parallelization on high-performance computing clusters.
• Seurat-based Clustering: Performs repeated clustering on subsamples using Seurat for single-cell analysis.
• Parameter Sensitivity Analysis: Evaluates sensitivity of clustering outcomes to parameter choices such as number of principal components, k nearest neighbors, and resolution settings.
• Visualization Tools: Provides an R package with visualization functions to interpret clustering stability and compare clusters.

Scientific Applications:

• Cell Type Identification and Validation: Assists identification and validation of novel cell types in single-cell genomics by assessing robustness of transcriptionally defined clusters.
• Parameter Selection and Interpretation: Guides selection of optimal clustering parameters to improve reliability of biological interpretations from scRNAseq data.

Methodology:

Subsampling cells to create multiple datasets; repeating clustering on these subsamples using Seurat across varied parameter settings; estimating cluster stability via the Jaccard similarity index; and automating subsampling, repeated clustering, and parallel execution via a Snakemake workflow on high-performance computing clusters.