| name | folddisco |
| description | Folddisco — fast discontinuous structural motif search across the protein universe (Steinegger lab). Use when: (1) Searching catalytic triads / zinc fingers / other short motifs in AFDB/PDB/ESM30, (2) Motif-level annotation after Foldseek fold hits, (3) Mining conserved geometry in MAG protein structures. Upstream: https://github.com/steineggerlab/folddisco. Web: https://search.foldseek.com/folddisco. Route via homology-search / metagenomics-llm; complement foldseek (full-chain) and mmseqs2 (sequence).
|
| license | MIT |
| category | analysis-tools |
| tags | ["Folddisco","motif","structure-search","AFDB","Foldseek"] |
| upstream | https://github.com/steineggerlab/folddisco |
| stage | mining |
Folddisco
Upstream: steineggerlab/folddisco ·
Web: https://search.foldseek.com/folddisco
Citation
Kim, H., Kim, R. S., Mirdita, M., Yoon, J. & Steinegger, M. Structural motif
search across the protein universe with Folddisco. Nat. Biotechnol. (2026).
https://doi.org/10.1038/s41587-026-03162-9
See also docs/references.md.
Analytical thinking
Foldseek finds similar folds / chains; many functions depend on a small
discontinuous motif (catalytic residues, metal sites). Folddisco indexes
geometric hashes so motif queries scale to AFDB50 / ESM30 / PDB. In the
Steinegger “levels of conservation” framing: sequence/fold (mmseqs2 /
foldseek) → multimer (foldseek multimer) → motif (folddisco).
| vs | Prefer Folddisco when |
|---|
foldseek | Motif / active-site geometry, not global fold TM |
mmseqs2 | Residues are discontinuous or sequence-divergent |
| HMM / PROSITE | Have a 3D motif definition from a structure |
Caveats: motif length limits (long selections fail on the web UI); hits need
manual residue chemistry checks; pin index version (v2 *.tar.lz4 vs legacy).
How to run
folddisco query -i index/h_sapiens_folddisco -p query/4CHA.pdb -q B57,B102,C195 -t 6
folddisco query -i index/afdb50_v4_folddisco -p query.pdb -t 8 --skip-match
Web: https://search.foldseek.com/folddisco
Key parameters
| Setting | Notes |
|---|
-p / -q | Query PDB/mmCIF and residue list (B57,B102,C195) |
-i | Precomputed Folddisco index — pin release |
-d / -a | Distance (Å) / angle (°) prefilter sensitivity |
--skip-match | Geometric search without AA identity constraints |
--sampling-ratio | Speed/sensitivity trade-off for long queries |
Decision tree
Structure-aware annotation need?
├─ Global fold / chain → foldseek
├─ Complex / interface → foldseek easy-multimersearch
├─ Short discontinuous motif → folddisco
├─ Sequence only → mmseqs2 / plmsearch
└─ Gene neighborhood modules → spacedust
Related skills
foldseek · metagenomics-llm · mmseqs2 · spacedust · plmsearch ·
deepblast · microbial-mining · tool-selection
