| name | homology-search |
| description | Route all homology / similarity search paradigms for shotgun metagenomics, including the Steinegger/Söding stack (MMseqs2, HH-suite, Foldseek family, PLASS, MetaEuk, Petasearch), vector DB (ERAST), and genome warehouse search (Phylign, LexicMap). Use when choosing how to search proteins, genes, or genomes.
|
| license | MIT |
| category | orchestration |
| tags | ["homology","search","Steinegger","Foldseek","MMseqs2","vector-DB","routing"] |
| stage | function |
Homology search routing
Shotgun projects need many search paradigms. This hub chooses among them.
For LM dark-matter framing also see metagenomics-llm. For protein LMs see
protein-language-model. For trees see phylogenomics.
Steinegger / Söding search stack
| Role | Prefer |
|---|
| Sequence search / cluster / taxonomy | mmseqs2 |
| Profile HMM remote homology | hh-suite |
| Protein-level assembly from reads | plass |
| Eukaryotic gene call via homology | metaeuk |
| Structure compress/index | foldcomp |
| Structure search / multimer / ProstT5 | foldseek |
| Structural motifs | folddisco |
| Multiple structure alignment | foldmason |
| Peta-scale protein search | petasearch |
| Conserved gene neighborhoods | spacedust |
| Taxonomy (DNA/protein) | metabuli (separate taxonomy skill) |
Other paradigms
| Paradigm | Prefer |
|---|
| PLM embedding search | protein-language-model → plmsearch · deepblast |
| Vector DB + rerank | erast |
| Phylogeny-compressed genomes | phylign |
| Gene vs million genomes | lexicmap |
| Catalogue AA aligner | diamond |
Hard rules:
- Do not merge BLAST/MMseqs, HH-suite, Foldseek, PLM, and vector-DB tables unlabeled
- Pin software and database/index release
- Low pLDDT structures inflate Foldseek false positives
plass proteins ≠ binnable contigs; metaeuk is for eukaryotes
Decision tree
What are you searching / recovering?
│
├─ Proteins / ORFs
│ ├─ Close / catalogue-scale → mmseqs2 / diamond
│ ├─ Profile remote homology / MSA for folding → hh-suite
│ ├─ Structures or ProstT5 / AFDB → foldseek (± folddisco / foldmason / foldcomp)
│ ├─ Peta-scale protein targets → petasearch
│ ├─ Sequence-only PLM → protein-language-model (plmsearch / deepblast)
│ ├─ AA embeddings / ProstT5 → protein-language-model (esm / prostt5)
│ └─ Billion-scale vector retrieval → erast
├─ Recover more proteins from reads → plass
├─ Eukaryotic genes on contigs → metaeuk
├─ Genes / plasmids vs huge genome sets
│ ├─ Phylogeny-compressed 661k-style → phylign
│ └─ Million-genome nucleotide index → lexicmap
├─ Dark modules / neighborhoods → spacedust
└─ Unsure → this hub + tool-selection
Related skills
mmseqs2 · hh-suite · plass · metaeuk · foldcomp · foldseek ·
folddisco · foldmason · petasearch · spacedust ·
plmsearch · deepblast · erast · phylign · lexicmap · diamond ·
protein-language-model · metagenomics-llm · phylogenomics · unicore ·
tool-selection
