TEDBench/

ted

Paper | GitHub Large-scale, non-redundant benchmark for protein fold classification built from Encyclopedia of Domains (TED) annotations projected onto the Foldseek-clustered AlphaFold Database. This dataset was presented in the paper Protein Fold Classification at Scale: Benchmarking and Pretraining.

	Dataset statistics




	

Split Structures

Train 369,740

Val 46,217

Test 46,218

965 CATH topology (T-level) classes — rare topologies with fewer… See the full description on the dataset page: https://huggingface.co/datasets/TEDBench/ted.