GiniClust

GiniClust identifies rare cell types in single-cell RNA-seq data by using the Gini index to quantify gene expression variability and detect cells with unique expression profiles.

Key Features:

• Gini-index based feature selection: Employs the Gini index to quantify gene expression variability across single-cell RNA-seq datasets and distinguish rare from dominant cell populations.
• High sensitivity and specificity: Validation against benchmark datasets demonstrated high sensitivity and specificity for detecting rare cellular subtypes.
• Application to public datasets: Applied to public single-cell RNA-seq datasets to identify Zscan4-expressing cells in mouse embryonic stem cells and hemoglobin-expressing cells in mouse cortex and hippocampus.
• Detection in mixed populations: Detects small numbers of normal cells within predominantly cancerous cell populations.

Scientific Applications:

• Developmental biology: Enables detection of rare developmental cell states such as Zscan4-expressing cells.
• Disease progression: Aids investigation of cellular heterogeneity relevant to disease progression.
• Tissue organization: Provides insights into tissue organization by uncovering rare cell types in heterogeneous tissues.
• Cancer research: Distinguishes normal and malignant cells within tumor microenvironments to inform analyses of tumor heterogeneity.

Methodology:

Computes the Gini index on single-cell RNA-seq gene expression to quantify variability and identifies cells with unique expression profiles; validated against benchmark datasets for sensitivity and specificity.