poem

poem implements a comprehensive set of external and internal evaluation metrics for assessing subpopulation detection in single-cell and spatial omics data.

Key Features:

• Clustering Evaluation Metrics: Provides metrics for assessing partitions or fuzzy partitions derived from clustering results.
• Embedding and Graph Evaluation Metrics: Provides metrics for evaluating subpopulation identification results within embeddings or graph representations.
• Spatial Domain Detection Metrics: Provides metrics for comparing spatial domain detection results against ground truth labels, including novel external metrics tailored to spatially aware measurements.
• Spatial Error Visualization: Provides tools for visualizing spatial errors.

Scientific Applications:

• Single-Cell Omics Analysis: Supports evaluation of cell embeddings, graph construction, and clustering in single-cell data analysis workflows.
• Spatial Omics Domain Detection: Enables systematic validation of spatial domain detection methods in spatially resolved transcriptomics data.
• Metric Selection and Benchmarking: Provides a framework for understanding metric behavior, biological relevance, and potential biases to guide selection of appropriate validation metrics.

Methodology:

The package implements existing metrics and develops novel metrics, including new external metrics tailored to spatially aware measurements for spatial omics data.