SCISSORS

SCISSORS identifies rare cell types and subclusters in single-cell RNA sequencing (scRNA-seq) data by semi-supervised optimization of rare-cell silhouettes to improve detection of low-abundance populations within heterogeneous samples.

Key Features:

• Integration with Seurat R package: Integrates into Seurat-based scRNA-seq analysis workflows to operate on existing single-cell datasets and objects.
• Silhouette scoring: Employs silhouette scoring to estimate heterogeneity within clusters and to quantify cluster cohesiveness and separability.
• Semi-supervised reclustering: Uses a multi-step semi-supervised reclustering process to reveal rare cells within heterogeneous clusters.
• Marker gene identification: Identifies marker genes with high specificity for detected subclusters to support characterization of low-abundance cell populations.

Scientific Applications:

• Developmental biology: Resolves granular cellular heterogeneity and detects transient or low-abundance cell states in developmental scRNA-seq datasets.
• Cancer research: Identifies rare tumor subpopulations or microenvironmental cell types that may be masked in bulk clustering of scRNA-seq data.
• Immunology: Detects low-frequency immune cell subsets and refines subcluster definitions in heterogeneous immune profiling by scRNA-seq.

Methodology:

Integrates with the Seurat R package, applies silhouette scoring to estimate within-cluster heterogeneity, performs multi-step semi-supervised reclustering to reveal rare subclusters, and identifies marker genes specific to those subclusters.