Cryo-EM Map Post-Processing: Curated Average Maps and Metadata

Andreea V. Florea Toma and colleagues released this curated collection in 2026 to provide a standardized benchmark for cryo-EM map post-processing. The dataset contains average maps and associated metadata partitioned into non-overlapping training, validation, and test subsets for reproducible method evaluation.

Use Cases

• Training neural networks for map sharpening using 'average maps' as the primary input
• Evaluating post-processing accuracy by comparing results against 'Q-scores' and 'Phenix' metrics
• Standardizing structural biology workflows using the 'non-overlapping' data partitions to ensure unbiased testing

Strengths

• Standardized partitioning into non-overlapping training, validation, and test subsets
• Includes precomputed quality metrics derived from Phenix tools and Q-scores
• Provides a unified evaluation framework for reproducible benchmarking

Limitations

• Requires integration with multiple external Dataverse datasets to access full reference data
• Specific record counts and file sizes are not provided in the primary metadata

Provenance

Source

Andreea V. Florea Toma, Dataverse

Collection Method

Curated from existing cryo-EM data and processed through a unified evaluation framework

Freshness

Last updated March 2026.