DENet: Protein Fitness and Directed Evolution Data for Deep Learning

George Song provides datasets and auxiliary files for reproducing analyses from a paper on navigating protein fitness landscapes via deep learning. The 1.9 GB repository includes protein sequences, fitness scores, mutant libraries, and structure files. It was last updated on 2026-05-11 under a CC-BY-4.0 license.

Use Cases

• Training deep learning models on directed evolution trajectories based on the provided sequence and fitness data.
• Benchmarking protein fitness prediction algorithms using the included mutational-scanning datasets.
• Analyzing protein structure-function relationships based on the processed contact-map and PDB files.
• Studying co-mutation patterns in protein evolution using the generated co-mutation files.

Strengths

• Repository size is 1.9 GB, indicating substantial supporting material.
• Includes multiple data types such as protein sequences, fitness scores, and structural files.
• Supports reproducibility with files structured to match an accompanying GitHub repository.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect temporal or source bias inherent to the original research project.

Provenance

Source

George Song via figshare.

Collection Method

Created to support analyses for a specific research paper.

Freshness

Last updated 2026-05-11 01:47:29; freshness should be verified.