Benchmarking Data for Protein Structure Similarity Search Using Embeddings and Established Methods

This dataset contains benchmarking results for evaluating protein structure similarity search using embedding-based methods in comparison with established tools such as TM-align, DALI, and Foldseek. Results are provided for multiple benchmarks, including domain-level (SCOPe40 v2.01), full-length chains (PDB non-redundant set), computed structure models (AlphaFold DB multidomain proteins), and multimeric assemblies (3DComplex). These data support the analyses presented in "Multi-scale structural similarity embedding search across entire proteomes.", and enable independent verification or reuse in related studies of protein structural search and comparison.