PDB Search inputs and parameters
Input formats, parameters, and validation for PDB structure search and download.
You provide either a free-text query to search the database or a 4-character PDB ID to download a specific structure — the orchestrator picks which from your phrasing. When a search returns several candidates, resolution is the input that most affects downstream quality: it is the primary criterion for choosing which structure to download.
Input methods
PDB has two operations on the platform: a search of the structure database and a download of a specific structure. Both run when you ask Ivy in chat; there is no form, job, or command for them.
Chat input (natural language)
Type a structure request in plain English. The orchestrator determines whether to search or download.
Search examples:
Download examples:
What you can ask for
PDB search
| You can set | Default | Limits |
|---|---|---|
| Query | Required | Protein name, organism, technique, ligand, or any keyword |
| Number of results | 25 | 1-100 |
The search performs a full-text query against all PDB metadata fields including title, organism, experimental method, authors, and deposited ligand names.
PDB download
| You can set | Default | Limits |
|---|---|---|
| PDB ID | Required | A 4-character PDB identifier (e.g., 1M17, 7BQY) |
| File format | PDB | PDB or mmCIF |
Search query strategies
Protein name search
Search by protein name, gene name, or common names:
Organism-specific search
Include the organism name in your query:
Method-specific search
Filter by experimental method:
Ligand-based search
Find structures with specific bound ligands:
Author search
Find structures deposited by specific researchers:
PDB ID format
PDB identifiers follow a strict format:
- Length: Exactly 4 characters
- Pattern: Starts with a digit (0-9), followed by 3 alphanumeric characters
- Examples: 1M17, 7BQY, 3POZ, 8DFN
- Case: Case-insensitive (1m17 and 1M17 are equivalent; the platform converts to uppercase)
Extended PDB IDs
The PDB is transitioning to extended identifiers (PDB-Dev format) for future entries. Currently, all active entries use the 4-character format.
File format options
PDB format (default)
- Extension:
.pdb - Description: Legacy text-based format, widely supported by molecular viewers and analysis tools
- Best for: DiffDock input, quick visualization, legacy tool compatibility
- Limitations: Cannot represent very large structures (>99,999 atoms), limited precision for coordinates
mmCIF format
- Extension:
.cif - Description: Macromolecular Crystallographic Information File, the modern replacement for PDB format
- Best for: Large structures, multi-model entries, complete metadata, new depositions
- Advantages: No atom count limits, richer metadata, mandatory for new PDB depositions
Input validation
Search query validation
- Minimum: Query must be a non-empty string
- Special characters: Most characters are passed through to the RCSB Search API
- Empty queries: Rejected with an error message
PDB ID validation
- Format: Must match the pattern: one digit followed by three alphanumeric characters
- Regex pattern:
^[0-9][A-Za-z0-9]{3}$ - Invalid IDs: Return an error. Common mistakes include using UniProt accessions (6 characters) instead of PDB IDs (4 characters).
- Non-existent IDs: Valid format but no structure exists. Returns a download error.
Limit validation
- Minimum: 1
- Maximum: 100
- Default: 25
- Values outside range: Clamped to nearest boundary
Example prompts
Search for high-resolution kinase structures
Download a specific structure in PDB format
Download a structure in mmCIF format
Search for structures with bound drugs
Resolution guidelines for structure selection
When search results return multiple structures, resolution is the primary quality criterion for selecting which structure to download:
| Resolution | Recommendation |
|---|---|
| <1.5 A | Excellent. Atomic-level detail, ideal for any analysis. |
| 1.5-2.0 A | Very good. Suitable for docking, binding site analysis, and detailed modeling. |
| 2.0-2.5 A | Good. Adequate for most structure-based drug design work. |
| 2.5-3.0 A | Moderate. Backbone reliable, but side chain positions may be uncertain. |
| 3.0-3.5 A | Low. Useful for overall fold and domain organization. Caution for detailed work. |
| >3.5 A | Very low. Limited to general topology. Not recommended for docking or binding analysis. |
NMR structures do not have a resolution metric. Their quality is assessed by other metrics (ensemble RMSD, Ramachandran statistics).
Read next
Was this page helpful?