SAME-clustering

SAME-clustering performs ensemble clustering of single-cell RNA-seq (scRNA-seq) data by aggregating multiple clustering solutions using a mixture model to improve cluster assignments and estimate the number of clusters.

Key Features:

• Mixture model-based ensemble: Aggregates multiple clustering outputs via a mixture model to produce a consensus clustering solution.
• Diverse-subset selection: Evaluates individual clustering solutions and selects a maximally diverse subset for combination.
• Improved clustering accuracy: Demonstrated superior clustering performance compared to standalone methods across evaluated datasets.
• Benchmarking across datasets: Validated on 15 scRNA-seq datasets with cluster counts ranging from 3 to 15 and cell counts from 49 to 32,695.
• Platform robustness: Applied successfully to datasets generated by various scRNA-seq platforms.
• Broad applicability: Ensemble methodology is applicable to clustering problems beyond scRNA-seq.

Scientific Applications:

• Cell type identification: Identifies distinct cell types and their transcriptomic signatures within scRNA-seq data to elucidate tissue complexity.
• Characterization of cellular heterogeneity: Supports analysis of cellular heterogeneity and dynamic processes, including investigations relevant to disease mechanisms.
• General clustering tasks: Provides an ensemble framework applicable to diverse clustering challenges beyond single-cell transcriptomics.

Methodology:

SAME-clustering generates individual clustering solutions using different methods, evaluates solution diversity to select a maximally diverse subset, and combines the selected solutions via a mixture model ensemble.