Sub-Cluster Identification through Semi-Supervised Optimization of Rare-Cell Silhouettes (SCISSORS) in single-cell...

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) allows simultaneous molecular profiling of numerous cells in complex biological samples.
• Current methods often rely on unsupervised clustering, which can be sensitive to parameter choices, potentially misidentifying cell types.
• Identifying rare cell types is challenging due to their subtle gene expression patterns and limited influence on overall data.

Purpose of the Study:

• To present SCISSORS, a novel framework for accurate subclustering and rare cell type identification in scRNA-seq data.
• To enhance the detection of low-abundance cell populations often missed by conventional clustering approaches.
• To provide a method for identifying highly specific marker genes for identified cell types.

Main Methods:

• SCISSORS utilizes silhouette scoring to estimate cluster heterogeneity.
• A multi-step semi-supervised reclustering process is employed to reveal rare cells within heterogeneous clusters.
• The framework is integrated with the Seurat R package for seamless incorporation into existing pipelines.

Main Results:

• SCISSORS accurately profiles subclusters, enabling precise identification of rare cell types.
• The method effectively detects cells of extremely low abundance.
• SCISSORS facilitates the identification of highly specific marker genes for distinct cell populations.

Conclusions:

• SCISSORS offers a robust solution for rare cell type identification in scRNA-seq data.
• The framework improves upon existing methods by addressing the limitations of unsupervised clustering for rare cell detection.
• SCISSORS enhances the analytical capabilities for exploring cellular heterogeneity in complex biological systems.