MarcoPolo

MarcoPolo identifies differentially expressed genes in single-cell RNA sequencing (scRNA-seq) data without relying on prior cell clustering.

Key Features:

• Independence from Clustering: Identifies marker genes without initial cell clustering, avoiding biases from clustering parameters and manual decisions.
• Bimodal Distribution Evaluation: Assesses whether gene expression distributions are bimodal as an indicator of differential expression across cell populations.
• Cross-Gene Pattern Recognition: Evaluates whether similar expression patterns are observed in other genes to enhance DEG identification robustness.
• Spatial Proximity Analysis: Considers whether expressing cells are proximal in a low-dimensional embedding to provide spatial context for expression differences.
• Output Format: Produces analysis results in an HTML file format.
• Empirical Validation: Validated on real datasets with fluorescence-activated cell sorting (FACS)-purified cell labels and shown to recover marker genes effectively compared to existing methods.

Scientific Applications:

• Marker gene discovery in scRNA-seq: Detects differentially expressed genes and marker genes that can be missed by clustering-dependent pipelines.
• Resolving subtle cell-type differences: Reveals subtle transcriptional differences between closely related cell types, for example anterior versus mid primitive streak cells.
• Analysis when clustering is unreliable: Applies to datasets where shared global structures or parameter sensitivity make clustering-based identification of cell types unreliable.

Methodology:

Performs systematic, clustering-free evaluation of gene expression by statistically assessing bimodality of gene distributions, detecting similar expression patterns across genes, and assessing spatial proximity of expressing cells in low-dimensional representations.