triku

triku prioritizes genes that define primary cell populations in single-cell RNA sequencing (scRNA-seq) datasets by selecting features expressed by clusters of neighboring cells in a nearest neighbor graph.

Key Features:

• Feature selection for scRNA-seq: Enhances feature selection specifically for single-cell RNA sequencing datasets to identify informative genes.
• Nearest neighbor graph-based selection: Identifies and selects genes expressed by clusters of cells that are spatially proximate in a nearest neighbor graph.
• Mutual information and silhouette scoring: Evaluates genes using mutual information and silhouette coefficient measurements to assess informativeness and clustering quality.
• Primary population marker prioritization: Prioritizes genes that are instrumental in defining primary cell populations and distinguishing cell types or states.
• Benchmarking recovery: Demonstrates accurate recovery of cell populations across artificial and real-world benchmarking scenarios.
• Functional enrichment association: Produces gene sets that are more likely to associate with relevant Gene Ontology terms.
• Reduced technical gene content: Selected gene sets exhibit reduced representation of ribosomal and mitochondrial genes.

Scientific Applications:

• Cell-type identification: Identification and prioritization of marker genes for primary cell populations in scRNA-seq studies.
• Clustering improvement: Improving clustering quality by selecting features that enhance separation as measured by silhouette coefficient and mutual information.
• Functional interpretation: Generation of gene sets suitable for downstream Gene Ontology-based functional enrichment and biological interpretation while minimizing ribosomal and mitochondrial confounders.

Methodology:

Selection uses a nearest neighbor graph to find clusters of proximate cells and evaluates gene expression patterns using mutual information and silhouette coefficient measurements to rank and choose informative genes.