foldseek

Name: foldseek
Author: adaptyvbio

35installs20forks

How to install foldseek?

adaptyvbio/protein-design-skills/foldseek

$npx -y skills add adaptyvbio/protein-design-skills --skill foldseek

Installs into the current project.

Summary

Structure similarity search with Foldseek. Use this skill when: (1) Finding similar structures in PDB/AFDB databases, (2) Structural homology search, (3) Database queries by 3D structure, (4) Finding remote homologs not detected by sequence, (5) Clustering structures by similarit

Files · 1

View on GitHub

SKILL.md

1	# Foldseek Structure Search
2
3	## Prerequisites
4
5	\| Requirement \| Minimum \| Recommended \|
6	\|-------------\|---------\|-------------\|
7	\| Python \| 3.8+ \| 3.10 \|
8	\| RAM \| 8GB \| 16GB \|
9	\| Disk \| 10GB \| 50GB (for local databases) \|
10
11	## How to run
12
13	Note: Foldseek can run locally or via web server. No GPU required.
14
15	### Option 1: Web Server (Quick; rate-limited, use sparingly)
16	```bash
17	# Upload structure to web server
18	curl -X POST "https://search.foldseek.com/api/ticket" \
19	-F "q=@query.pdb" \
20	-F "database[]=afdb50" \
21	-F "database[]=pdb100"
22	```
23
24	### Option 2: Local installation
25	```bash
26	# Install Foldseek
27	conda install -c conda-forge -c bioconda foldseek
28
29	# Search PDB
30	foldseek easy-search query.pdb /path/to/pdb100 results.m8 tmp/
31
32	# Search AlphaFold DB
33	foldseek easy-search query.pdb /path/to/afdb50 results.m8 tmp/
34	```
35
36	### Option 3: Python API
37	```python
38	import subprocess
39	import pandas as pd
40
41	def foldseek_search(query_pdb, database, output="results.m8"):
42	"""Run Foldseek search."""
43	subprocess.run([
44	"foldseek", "easy-search",
45	query_pdb, database, output, "tmp/",
46	"--format-output", "query,target,pident,alnlen,evalue,bits"
47	])
48	return pd.read_csv(output, sep="\t",
49	names=["query", "target", "pident", "alnlen", "evalue", "bits"])
50	```
51
52	## Key parameters
53
54	\| Parameter \| Default \| Description \|
55	\|-----------\|---------\|-------------\|
56	\| `--min-seq-id` \| 0.0 \| Minimum sequence identity \|
57	\| `-e` \| 0.001 \| E-value threshold \|
58	\| `--alignment-type` \| 2 \| 0=3Di, 1=TM, 2=3Di+AA \|
59	\| `--max-seqs` \| 1000 \| Max hits to pass through prefilter; reducing this affects sensitivity \|
60
61	## Databases
62
63	\| Database \| Description \| Size \|
64	\|----------\|-------------\|------\|
65	\| `pdb100` \| PDB chains \| ~340K structures \|
66	\| `afdb50` \| AlphaFold DB clustered at 50% sequence identity \| ~53M structures \|
67	\| `swissprot` \| SwissProt structures \| ~540K structures \|
68	\| `cath50` \| CATH domains \| ~50K domains \|
69
70	## Output format
71
72	```
73	# results.m8 (tabular)
74	query target pident alnlen evalue bits
75	query 1abc_A 85.2 120 1e-45 180.5
76	query 2def_B 72.1 115 1e-32 145.2
77	```
78
79	## Sample output
80
81	### Successful run
82	```
83	$ foldseek easy-search query.pdb pdb100 results.m8 tmp/
84	# results.m8 columns: query target pident alnlen mismatch gapopen qstart qend tstart tend evalue bits
85	query 1abc_A 85.2 120 ... 1e-45 180.5
86	query 2def_B 72.1 115 ... 1e-32 145.2
87	```
88
89	Hit identities and E-values above are placeholders; foldseek does not print the
90	`[INFO]` lines shown by some other tools.
91
92	## Decision tree
93
94	```
95	Should I use Foldseek?
96	│
97	├─ What are you searching?
98	│ ├─ By 3D structure → Foldseek ✓
99	│ ├─ By sequence → Use BLAST (uniprot skill)
100	│ └─ Both → Run both, compare results
101	│
102	└─ What do you need?
103	├─ Find structural homologs → Foldseek ✓
104	├─ Remote homolog detection → Foldseek ✓
105	├─ Structural clustering → Foldseek ✓
106	└─ Functional annotation → Cross-reference with UniProt
107	```
108
109	## Common use cases
110
111	### Find similar designs
112	```bash
113	# Compare your design to PDB
114	foldseek easy-search design.pdb pdb100 similar_natural.m8 tmp/
115	```
116
117	### Novelty check
118	```bash
119	# Ensure design is novel (low similarity to known)
120	foldseek easy-search design.pdb afdb50 novelty.m8 tmp/
121
122	# Novel if: top hit identity < 30%
123	```
124
125	### Scaffold search
126	```bash
127	# Find scaffolds for motif grafting
128	foldseek easy-search motif.pdb pdb100 scaffolds.m8 tmp/ \
129	--min-seq-id 0.0 -e 10
130	```
131
132	---
133
134	## Verify
135
136	```bash
137	wc -l results.m8 # Number of hits
138	```
139
140	---
141
142	## Troubleshooting
143
144	No hits: Lower e-value threshold, try larger database
145	Too many hits: Increase min-seq-id threshold
146	Slow search: Use smaller database
147
148	### Error interpretation
149
150	\| Error \| Cause \| Fix \|
151	\|-------\|-------\|-----\|
152	\| `Database not found` \| Wrong path \| Check database location \|
153	\| `Invalid PDB` \| Malformed structure \| Validate PDB format \|
154	\| `Out of memory` \| Large database \| Use more RAM or web server \|
155
156	---
157
158	Next: Download hits with `pdb` skill → use for scaffold design.

People who liked this also viewed

6 picks

Type

foldseek

35installs20forks

View on GitHub

How to install foldseek?

adaptyvbio/protein-design-skills/foldseek

$npx -y skills add adaptyvbio/protein-design-skills --skill foldseek

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Foldseek Structure Search
2
3	## Prerequisites
4
5	\| Requirement \| Minimum \| Recommended \|
6	\|-------------\|---------\|-------------\|
7	\| Python \| 3.8+ \| 3.10 \|
8	\| RAM \| 8GB \| 16GB \|
9	\| Disk \| 10GB \| 50GB (for local databases) \|
10
11	## How to run
12
13	Note: Foldseek can run locally or via web server. No GPU required.
14
15	### Option 1: Web Server (Quick; rate-limited, use sparingly)
16	```bash
17	# Upload structure to web server
18	curl -X POST "https://search.foldseek.com/api/ticket" \
19	-F "q=@query.pdb" \
20	-F "database[]=afdb50" \
21	-F "database[]=pdb100"
22	```
23
24	### Option 2: Local installation
25	```bash
26	# Install Foldseek
27	conda install -c conda-forge -c bioconda foldseek
28
29	# Search PDB
30	foldseek easy-search query.pdb /path/to/pdb100 results.m8 tmp/
31
32	# Search AlphaFold DB
33	foldseek easy-search query.pdb /path/to/afdb50 results.m8 tmp/
34	```
35
36	### Option 3: Python API
37	```python
38	import subprocess
39	import pandas as pd
40
41	def foldseek_search(query_pdb, database, output="results.m8"):
42	"""Run Foldseek search."""
43	subprocess.run([
44	"foldseek", "easy-search",
45	query_pdb, database, output, "tmp/",
46	"--format-output", "query,target,pident,alnlen,evalue,bits"
47	])
48	return pd.read_csv(output, sep="\t",
49	names=["query", "target", "pident", "alnlen", "evalue", "bits"])
50	```
51
52	## Key parameters
53
54	\| Parameter \| Default \| Description \|
55	\|-----------\|---------\|-------------\|
56	\| `--min-seq-id` \| 0.0 \| Minimum sequence identity \|
57	\| `-e` \| 0.001 \| E-value threshold \|
58	\| `--alignment-type` \| 2 \| 0=3Di, 1=TM, 2=3Di+AA \|
59	\| `--max-seqs` \| 1000 \| Max hits to pass through prefilter; reducing this affects sensitivity \|
60
61	## Databases
62
63	\| Database \| Description \| Size \|
64	\|----------\|-------------\|------\|
65	\| `pdb100` \| PDB chains \| ~340K structures \|
66	\| `afdb50` \| AlphaFold DB clustered at 50% sequence identity \| ~53M structures \|
67	\| `swissprot` \| SwissProt structures \| ~540K structures \|
68	\| `cath50` \| CATH domains \| ~50K domains \|
69
70	## Output format
71
72	```
73	# results.m8 (tabular)
74	query target pident alnlen evalue bits
75	query 1abc_A 85.2 120 1e-45 180.5
76	query 2def_B 72.1 115 1e-32 145.2
77	```
78
79	## Sample output
80
81	### Successful run
82	```
83	$ foldseek easy-search query.pdb pdb100 results.m8 tmp/
84	# results.m8 columns: query target pident alnlen mismatch gapopen qstart qend tstart tend evalue bits
85	query 1abc_A 85.2 120 ... 1e-45 180.5
86	query 2def_B 72.1 115 ... 1e-32 145.2
87	```
88
89	Hit identities and E-values above are placeholders; foldseek does not print the
90	`[INFO]` lines shown by some other tools.
91
92	## Decision tree
93
94	```
95	Should I use Foldseek?
96	│
97	├─ What are you searching?
98	│ ├─ By 3D structure → Foldseek ✓
99	│ ├─ By sequence → Use BLAST (uniprot skill)
100	│ └─ Both → Run both, compare results
101	│
102	└─ What do you need?
103	├─ Find structural homologs → Foldseek ✓
104	├─ Remote homolog detection → Foldseek ✓
105	├─ Structural clustering → Foldseek ✓
106	└─ Functional annotation → Cross-reference with UniProt
107	```
108
109	## Common use cases
110
111	### Find similar designs
112	```bash
113	# Compare your design to PDB
114	foldseek easy-search design.pdb pdb100 similar_natural.m8 tmp/
115	```
116
117	### Novelty check
118	```bash
119	# Ensure design is novel (low similarity to known)
120	foldseek easy-search design.pdb afdb50 novelty.m8 tmp/
121
122	# Novel if: top hit identity < 30%
123	```
124
125	### Scaffold search
126	```bash
127	# Find scaffolds for motif grafting
128	foldseek easy-search motif.pdb pdb100 scaffolds.m8 tmp/ \
129	--min-seq-id 0.0 -e 10
130	```
131
132	---
133
134	## Verify
135
136	```bash
137	wc -l results.m8 # Number of hits
138	```
139
140	---
141
142	## Troubleshooting
143
144	No hits: Lower e-value threshold, try larger database
145	Too many hits: Increase min-seq-id threshold
146	Slow search: Use smaller database
147
148	### Error interpretation
149
150	\| Error \| Cause \| Fix \|
151	\|-------\|-------\|-----\|
152	\| `Database not found` \| Wrong path \| Check database location \|
153	\| `Invalid PDB` \| Malformed structure \| Validate PDB format \|
154	\| `Out of memory` \| Large database \| Use more RAM or web server \|
155
156	---
157
158	Next: Download hits with `pdb` skill → use for scaffold design.

People who liked this also viewed

6 picks

Type