PicturedRocks

PicturedRocks applies information-theoretic feature selection to single-cell RNA-sequencing (scRNA-seq) data to identify small, informative marker gene sets that discriminate cell types within complex cellular mixtures.

Key Features:

• Information-Theoretic Feature Selection: Implements information-theoretic approaches capable of handling binary or multiclass and univariate or multivariate data to identify informative genes.
• Small Efficient Marker Sets: Focuses on selecting small and efficient sets of marker genes that together provide discriminatory power for cell type identification.
• Benchmarking Against Differential Expression: Benchmarks differential expression methods against information-theoretic selection, showing that information-theoretic methods can identify genes as informative as or more than traditional approaches in some datasets.
• Python Implementation and scanpy Compatibility: Implemented in Python with compatibility for integration into scanpy-based scRNA-seq workflows.

Scientific Applications:

• Marker gene selection for scRNA-seq: Identifies marker genes for distinguishing cell types in single-cell RNA-seq studies.
• Discovery of cellular populations and hierarchies: Supports uncovering novel cellular populations and hierarchical relationships within biological systems.
• Evaluation of differential expression methods: Provides a framework to evaluate and compare the informativeness of genes selected by differential expression versus information-theoretic methods.

Methodology:

Performs theoretical analysis and application of information-theoretic feature selection algorithms to select marker genes based on their informativeness for cell type discrimination.