SNN-Cliq

SNN-Cliq performs clustering of single-cell transcriptome gene expression data using shared nearest neighbor graph methods to identify cell types from high-dimensional, noisy single-cell datasets.

Key Features:

• Shared Nearest Neighbor (SNN) Concept: Leverages the shared nearest neighbor approach to define similarities between individual cells in high-dimensional gene expression space, mitigating issues from the curse of dimensionality.
• Graph Theory-Based Clustering Algorithm: Constructs a graph with nodes representing cells and edges encoding SNN relationships and applies graph-theory clustering to group cells by transcriptomic profiles.
• High Accuracy in Reflecting Cell Types: Produces clusters that align with actual cell type origins and has been demonstrated to outperform other methods on synthetic and real experimental datasets.
• Implementations: Provided implementations in MATLAB and Python.

Scientific Applications:

• Single-cell transcriptome analysis: Addresses clustering of noisy, high-dimensional datasets with many genes and comparatively few cells.
• Characterization of cellular heterogeneity: Groups cells by gene expression to reveal functional variation within cell populations.
• Applications in developmental biology, cancer research, and immunology: Enables identification of distinct cell types and states relevant to these fields.

Methodology:

Computes pairwise cell similarities using the shared nearest neighbor (SNN) approach, constructs a cell graph with SNN-based edges, and applies a graph theory-based clustering algorithm.