alphafold_database

Name: alphafold_database
Author: ai4protein

244stars35forks

How to install alphafold_database?

ai4protein/venusfactory2/alphafold_database

$npx -y skills add ai4protein/VenusFactory2 --skill alphafold_database

Installs into the current project.

Summary

Access AlphaFold 200M+ AI-predicted protein structures. Retrieve structures by UniProt ID, download PDB/mmCIF files, analyze confidence metrics (pLDDT, PAE), for drug discovery and structural biology.

Files · 1

View on GitHub

SKILL.md

1	# AlphaFold Database
2
3	## Overview
4
5	AlphaFold DB is a public repository of AI-predicted 3D protein structures for over 200 million proteins, maintained by DeepMind and EMBL-EBI. Access structure predictions with confidence metrics, download coordinate files, retrieve bulk datasets, and integrate predictions into computational workflows.
6
7	## Project Tools (VenusFactory2)
8
9	The following agent tools are exposed for AlphaFold in this codebase. Query tools that would return structure/metadata text into the conversation are not exposed, because they would put large PDB/mmCIF or JSON payloads into the context and cause context explosion. Use the download tools instead: they write to disk and return only `{success, file_path}` in JSON. The two `analyze_*` tools then read those local files and emit a small structured report — keeping the analysis off the wire.
10
11	\| Tool \| Args \| Returns \| Description \|
12	\|------\|------\|---------\|--------------\|
13	\| download_alphafold_structure_by_uniprot_id \| `uniprot_id` (required), `out_dir` (required), `format` (default `"pdb"`, choices `pdb` \\| `cif`), `version` (default `"v6"`, choices `v1` \\| `v2` \\| `v4` \\| `v6`), `fragment` (default `1`, ≥1) \| JSON: `{success: bool, file_path: str \\| null}` \| Download AlphaFold structure (PDB or mmCIF) to `out_dir`. `file_path` is the path to the downloaded file. \|
14	\| download_alphafold_metadata_by_uniprot_id \| `uniprot_id` (required), `out_dir` (required) \| JSON: `{success: bool, file_path: str \\| null}` \| Download AlphaFold prediction metadata (JSON) to `out_dir`. `file_path` is the path to the saved metadata file. \|
15	\| analyze_alphafold_plddt_by_metadata_file \| `metadata_path` (required, path to a downloaded AlphaFold metadata JSON) \| JSON: `{status, content, biological_metadata {uniprot_id, global_plddt, fractions {very_low, low, confident, very_high}, conclusion}}` \| Parse pLDDT fractions and emit a human-readable confidence verdict. Pure local analysis (no network). \|
16	\| analyze_alphafold_pae_by_pae_file \| `pae_path` (required, path to a downloaded PAE JSON file), `distance_cutoff` (default `7.0`), `min_domain_size` (default `40`) \| JSON: `{status, content, biological_metadata {matrix_shape, mean_pae, max_pae, min_pae, confident_pairs_pct, domains [{start, end, length}], conclusion}}` \| Walk the PAE matrix to detect rigid sub-domains, merge into global domains, report boundaries and aggregate stats. Pure local analysis. \|
17
18	- Always call with a valid `out_dir`; the download tools create the directory if needed.
19	- The analyze tools require a previously downloaded file path; chain them after the corresponding download tool.
20	- Structure URLs follow: `https://alphafold.ebi.ac.uk/files/{entry_id}-model_{version}.{pdb\|cif}` with `entry_id` = `AF-{uniprot_id}-F{fragment}`. PAE JSON URL: `…/{entry_id}-predicted_aligned_error_{version}.json`. See `references/api_reference.md` for full API details.
21
22	## When to Use This Skill
23
24	This skill should be used when working with AI-predicted protein structures in scenarios such as:
25
26	- Retrieving protein structure predictions by UniProt ID or protein name
27	- Downloading PDB/mmCIF coordinate files for structural analysis
28	- Analyzing prediction confidence metrics (pLDDT, PAE) to assess reliability
29	- Accessing bulk proteome datasets via Google Cloud Platform
30	- Comparing predicted structures with experimental data
31	- Performing structure-based drug discovery or protein engineering
32	- Building structural models for proteins lacking experimental structures
33	- Integrating AlphaFold predictions into computational pipelines
34
35	## Core Capabilities
36
37	### 1. Searching and Retrieving Predictions
38
39	Using Biopython (Recommended):
40
41	The Biopython library provides the simplest interface for retrieving AlphaFold structures:
42
43	```python
44	from Bio.PDB import alphafold_db
45
46	# Get all predictions for a UniProt accession
47	predictions = list(alphafold_db.get_predictions("P00520"))
48
49	# Download structure file (mmCIF format)
50	for prediction in predictions:
51	cif_file = alphafold_db.download_cif_for(prediction, directory="./structures")
52	print(f"Downloaded: {cif_file}")
53
54	# Get Structure objects directly
55	from Bio.PDB import MMCIFParser
56	structures = list(alphafold_db.get_structural_models_for("P00520"))
57	```
58
59	Direct API Access:
60
61	Query predictions using REST endpoints:
62
63	```python
64	import requests
65
66	# Get prediction metadata for a UniProt accession
67	uniprot_id = "P00520"
68	api_url = f"https://alphafold.ebi.ac.uk/api/prediction/{uniprot_id}"
69	response = requests.get(api_url)
70	prediction_data = response.json()
71
72	# Extract AlphaFold ID
73	alphafold_id = prediction_data[0]['entryId']
74	print(f"AlphaFold ID: {alphafold_id}")
75	```
76
77	**Using Uni

People who liked this also viewed

6 picks

Type

alphafold_database

244stars35forks

View on GitHub

How to install alphafold_database?

ai4protein/venusfactory2/alphafold_database

$npx -y skills add ai4protein/VenusFactory2 --skill alphafold_database

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# AlphaFold Database
2
3	## Overview
4
5	AlphaFold DB is a public repository of AI-predicted 3D protein structures for over 200 million proteins, maintained by DeepMind and EMBL-EBI. Access structure predictions with confidence metrics, download coordinate files, retrieve bulk datasets, and integrate predictions into computational workflows.
6
7	## Project Tools (VenusFactory2)
8
9	The following agent tools are exposed for AlphaFold in this codebase. Query tools that would return structure/metadata text into the conversation are not exposed, because they would put large PDB/mmCIF or JSON payloads into the context and cause context explosion. Use the download tools instead: they write to disk and return only `{success, file_path}` in JSON. The two `analyze_*` tools then read those local files and emit a small structured report — keeping the analysis off the wire.
10
11	\| Tool \| Args \| Returns \| Description \|
12	\|------\|------\|---------\|--------------\|
13	\| download_alphafold_structure_by_uniprot_id \| `uniprot_id` (required), `out_dir` (required), `format` (default `"pdb"`, choices `pdb` \\| `cif`), `version` (default `"v6"`, choices `v1` \\| `v2` \\| `v4` \\| `v6`), `fragment` (default `1`, ≥1) \| JSON: `{success: bool, file_path: str \\| null}` \| Download AlphaFold structure (PDB or mmCIF) to `out_dir`. `file_path` is the path to the downloaded file. \|
14	\| download_alphafold_metadata_by_uniprot_id \| `uniprot_id` (required), `out_dir` (required) \| JSON: `{success: bool, file_path: str \\| null}` \| Download AlphaFold prediction metadata (JSON) to `out_dir`. `file_path` is the path to the saved metadata file. \|
15	\| analyze_alphafold_plddt_by_metadata_file \| `metadata_path` (required, path to a downloaded AlphaFold metadata JSON) \| JSON: `{status, content, biological_metadata {uniprot_id, global_plddt, fractions {very_low, low, confident, very_high}, conclusion}}` \| Parse pLDDT fractions and emit a human-readable confidence verdict. Pure local analysis (no network). \|
16	\| analyze_alphafold_pae_by_pae_file \| `pae_path` (required, path to a downloaded PAE JSON file), `distance_cutoff` (default `7.0`), `min_domain_size` (default `40`) \| JSON: `{status, content, biological_metadata {matrix_shape, mean_pae, max_pae, min_pae, confident_pairs_pct, domains [{start, end, length}], conclusion}}` \| Walk the PAE matrix to detect rigid sub-domains, merge into global domains, report boundaries and aggregate stats. Pure local analysis. \|
17
18	- Always call with a valid `out_dir`; the download tools create the directory if needed.
19	- The analyze tools require a previously downloaded file path; chain them after the corresponding download tool.
20	- Structure URLs follow: `https://alphafold.ebi.ac.uk/files/{entry_id}-model_{version}.{pdb\|cif}` with `entry_id` = `AF-{uniprot_id}-F{fragment}`. PAE JSON URL: `…/{entry_id}-predicted_aligned_error_{version}.json`. See `references/api_reference.md` for full API details.
21
22	## When to Use This Skill
23
24	This skill should be used when working with AI-predicted protein structures in scenarios such as:
25
26	- Retrieving protein structure predictions by UniProt ID or protein name
27	- Downloading PDB/mmCIF coordinate files for structural analysis
28	- Analyzing prediction confidence metrics (pLDDT, PAE) to assess reliability
29	- Accessing bulk proteome datasets via Google Cloud Platform
30	- Comparing predicted structures with experimental data
31	- Performing structure-based drug discovery or protein engineering
32	- Building structural models for proteins lacking experimental structures
33	- Integrating AlphaFold predictions into computational pipelines
34
35	## Core Capabilities
36
37	### 1. Searching and Retrieving Predictions
38
39	Using Biopython (Recommended):
40
41	The Biopython library provides the simplest interface for retrieving AlphaFold structures:
42
43	```python
44	from Bio.PDB import alphafold_db
45
46	# Get all predictions for a UniProt accession
47	predictions = list(alphafold_db.get_predictions("P00520"))
48
49	# Download structure file (mmCIF format)
50	for prediction in predictions:
51	cif_file = alphafold_db.download_cif_for(prediction, directory="./structures")
52	print(f"Downloaded: {cif_file}")
53
54	# Get Structure objects directly
55	from Bio.PDB import MMCIFParser
56	structures = list(alphafold_db.get_structural_models_for("P00520"))
57	```
58
59	Direct API Access:
60
61	Query predictions using REST endpoints:
62
63	```python
64	import requests
65
66	# Get prediction metadata for a UniProt accession
67	uniprot_id = "P00520"
68	api_url = f"https://alphafold.ebi.ac.uk/api/prediction/{uniprot_id}"
69	response = requests.get(api_url)
70	prediction_data = response.json()
71
72	# Extract AlphaFold ID
73	alphafold_id = prediction_data[0]['entryId']
74	print(f"AlphaFold ID: {alphafold_id}")
75	```
76
77	**Using Uni

People who liked this also viewed

6 picks

Type