Overview
Foldseek, developed by the Steinegger lab, enables fast and sensitive comparisons of large protein structure sets. The README describes monomer and multimer searches (Foldseek-Multimer), clustering, and optional sequence-input searches powered by a language model that bypass the need for pre-existing structures. Alignment combines the 3Di structural alphabet with amino acid letters, and the software runs on CPU with optional GPU acceleration.
The project is GPL-3.0 licensed. Publications in Nature Biotechnology (2023, doi:10.1038/s41587-023-01773-0) and Nature Methods (2025) document the method and the multimer extension, and a public web server is available at search.foldseek.com.
Foldseek is a comparison and search tool, not a structure prediction method; results depend on the queried databases and scoring settings. It is complementary to the sequence-search suite MMseqs2 from the same lab.
Key Features
- Structure search and clustering for large protein structure sets, including multimer (Foldseek-Multimer) searches.
- Alignment modes combining the 3Di structural alphabet with amino acid information; CPU execution with optional GPU acceleration.
- Optional language-model-based search from sequence input without pre-existing structures.
Use Cases
- Search a predicted or experimental structure against large structure databases to find relatives.
- Cluster structure sets (for example predicted proteomes) to organize results of structure predictions.
How to Use
Foldseek can be run locally following the repository instructions or used through the public server at search.foldseek.com. The README documents alignment modes, database preparation and GPU options. See: https://github.com/steineggerlab/foldseek
Sources and verification
This entry describes capabilities documented by the linked project sources. Catalog edits are not evidence of a successful run, compatibility test or scientific performance evaluation.
Repository source reviewed for this catalog version:
Read the editorial and verification policy