---
title: PDB Search & Download on the platform
description: Search and retrieve protein structures from the RCSB Protein Data Bank.
tags:
  - pdb
---

PDB Search and Download retrieve experimentally determined 3D structures — solved by X-ray crystallography, cryo-EM, or NMR — from the RCSB Protein Data Bank. Reach for it when an experimental structure exists, especially one with a co-crystallized ligand: it is the highest-confidence structure you can get. If no experimental structure has been deposited, predict one instead — [fetch a known protein with AlphaFold](/docs/alphafold) or [fold a novel sequence with ESMFold](/docs/esmfold). See [what you can achieve](/docs/capabilities#structure) for the full set.

## What PDB Search does

PDB Search queries the **RCSB Protein Data Bank (PDB)**, the global archive for experimentally determined three-dimensional structures of biological macromolecules, growing by tens of thousands of new depositions each year through X-ray crystallography, cryo-electron microscopy (cryo-EM), NMR spectroscopy, and other methods. On the platform, PDB Search and PDB Download run when you ask Ivy in chat (there is no form or job for them); they call RCSB directly on CPU with no GPU requirements, returning results in 3-30 seconds.

The RCSB PDB is managed by the **Research Collaboratory for Structural Bioinformatics** and is the U.S. data center of the Worldwide Protein Data Bank (wwPDB). Every experimentally determined macromolecular structure published in a scientific journal is required to be deposited in the PDB.

## Citations

> Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE. The Protein Data Bank. *Nucleic Acids Research*. 2000;28(1):235-242. doi: [10.1093/nar/28.1.235](https://doi.org/10.1093/nar/28.1.235)

> Burley SK, Bhikadiya C, Bi C, Bittrich S, Chao H, Chen L, et al. RCSB Protein Data Bank (RCSB.org): delivery of experimentally-determined PDB structures alongside one million computed structure models of proteins from artificial intelligence/machine learning. *Nucleic Acids Research*. 2023;51(D1):D488-D508. doi: [10.1093/nar/gkac1077](https://doi.org/10.1093/nar/gkac1077)

## Key capabilities

- **Full-text search**: Find structures by protein name, organism, author, ligand name, or any keyword
- **Structure download**: Retrieve PDB or mmCIF format files by PDB ID
- **Metadata retrieval**: Get resolution, experimental method, organism, release date, chain information, and bound ligands
- **Resolution-based filtering**: Focus on high-resolution structures suitable for detailed analysis
- **Direct chaining**: Downloaded structures feed immediately into DiffDock, P2Rank, PDB2PQR, and MM-GBSA

## How it works technically

The platform runs two PDB operations, each wrapping a different RCSB API. Both run when you ask Ivy in chat:

### PDB Search

Uses the [RCSB Search API v2](https://search.rcsb.org/) to find structures matching a text query. The search process:

1. **Query submission**: The platform sends a full-text search query to `https://search.rcsb.org/rcsbsearch/v2/query` as a POST request with the query string.
2. **ID retrieval**: The API returns matching PDB IDs ranked by relevance, along with a total count.
3. **Metadata enrichment**: For the top 10 results, the platform fetches detailed metadata from the RCSB Data API (`https://data.rcsb.org/rest/v1/core/entry/{pdbId}`), extracting title, experimental method, resolution, organism, and release date.

### PDB Download

Downloads the actual structure file content from `https://files.rcsb.org/download/{pdbId}.{format}`. Supports PDB and mmCIF (CIF) formats.

### When to use PDB Search

### Primary use cases

**Structure-based drug design**
- Find high-resolution crystal structures of drug targets with bound ligands
- Identify co-crystal structures showing how existing drugs bind to a target
- Source structures for molecular docking with DiffDock
- Select the best available structure for binding site analysis with P2Rank

**Protein structure analysis**
- Retrieve experimental structures to compare with AI predictions (ESMFold, AlphaFold)
- Examine protein-protein interaction interfaces
- Analyze domain organization and quaternary structure
- Study conformational changes across different crystal forms

**Target validation**
- Confirm that structural data exists for a potential drug target
- Assess structural quality (resolution, completeness) for downstream computational work
- Identify available structures with different ligands or mutations
- Map binding sites and allosteric pockets

**Cryo-EM structure retrieval**
- Find large macromolecular complex structures determined by cryo-EM
- Retrieve viral spike proteins, ribosomes, and membrane protein complexes
- Access structures at resolutions increasingly comparable to X-ray crystallography

### PDB Search vs. AlphaFold vs. ESMFold

| Feature | PDB Search | AlphaFold Fetch | ESMFold |
|---------|-----------|-----------------|---------|
| Data source | Experimental structures | Pre-computed AI predictions | On-demand AI predictions |
| Coverage | Experimentally solved structures only | ~200M+ predicted structures | Any sequence |
| Quality | Experimental (varies by resolution) | AI-predicted (pLDDT confidence) | AI-predicted (pLDDT confidence) |
| Ligand information | Yes (co-crystallized ligands) | No | No |
| Speed | 3-30 seconds | 2-5 seconds | 10-60 seconds |
| Best for | High-resolution analysis, docking | Broad coverage, quick models | Novel/engineered sequences |

**Rule of thumb**: Use PDB when an experimental structure exists (especially with bound ligands). Use AlphaFold Fetch for quick access to predicted structures for well-known proteins. Use ESMFold when no existing structure is available.

### When to use alternatives

| Need | Use instead |
|------|-------------|
| AI-predicted structures | ESMFold (fresh prediction) or AlphaFold Fetch (pre-computed) |
| Protein sequence | UniProt Search |
| Protein function/annotations | UniProt Search |
| Structure comparison | Structural alignment tools |
| Small molecule structures | PubChem |

## Integration with the platform ecosystem

### Upstream data sources

PDB Search accepts input from:
- **Free-text chat queries**: Protein names, PDB IDs, or descriptive searches
- **UniProt results**: PDB cross-references from UniProt entries
- **BLAST results**: Homologous structures identified by sequence similarity
- **Entrez Search**: Structure records from NCBI's structure database

### Downstream applications

PDB structures feed into:
- **DiffDock**: Molecular docking using the experimental structure as the protein target
- **P2Rank**: Binding site prediction on the 3D structure
- **PDB2PQR**: Electrostatic calculations and protonation state assignment
- **MM-GBSA**: Binding free energy rescoring for docked poses
- **Mol* viewer**: Interactive 3D visualization on the canvas

### Workflow chaining examples

**Structure-to-docking pipeline:**
1. PDB Search for "EGFR kinase high resolution"
2. PDB Download for the best structure (e.g., 1M17)
3. DiffDock for molecular docking with a candidate ligand
4. MM-GBSA for complex minimization and an energy report (not a binding affinity)

**Target characterization pipeline:**
1. UniProt Search for protein annotations
2. PDB Search for experimental structures (via cross-references)
3. P2Rank for binding site prediction
4. DiffDock for ligand docking

## Scope and limitations

### Coverage

- **Total structures**: growing continuously; check [RCSB PDB statistics](https://www.rcsb.org/stats/growth/growth-released-structures) for the current count
- **Growth rate**: tens of thousands of new structures deposited per year
- **Methods**: X-ray crystallography (~85%), cryo-EM (~12%), NMR (~2%), other (~1%)
- **Organisms**: Structures from thousands of organisms, with human and common model organisms best represented

### Limitations

- **Coverage gaps**: Not all proteins have experimental structures. The PDB covers a fraction of the known protein universe. For proteins without PDB entries, use ESMFold or AlphaFold Fetch.
- **Resolution variability**: Structure quality varies widely. Some entries have excellent resolution (<1.5 A) while others are low resolution (>4.0 A). Always check resolution before using a structure for detailed analysis.
- **Static snapshots**: Crystal structures represent a single conformational state. Proteins in solution may adopt different conformations relevant to drug binding.
- **Crystal packing artifacts**: Some structural features may be influenced by crystal contacts rather than biological interactions.
- **Metadata search only**: PDB Search performs full-text search. Structured queries (by resolution range, method, ligand presence, etc.) are not supported; ask Ivy to filter the returned hits by the resolution and method that come back with each one.
- **Top 10 enrichment**: For performance, metadata is fetched for only the top 10 search results. Additional PDB IDs are listed without detailed metadata.

### Performance characteristics

- **Search runtime**: 3-10 seconds for text search, plus 1-2 seconds per entry for metadata enrichment
- **Download runtime**: 2-5 seconds for structure file retrieval
- **Maximum search results**: Up to 100 PDB IDs returned, with full metadata for the top 10
- **File sizes**: PDB files typically range from 100 KB to 10 MB depending on structure size

## References

- RCSB PDB: [https://www.rcsb.org/](https://www.rcsb.org/)
- RCSB Search API documentation: [https://search.rcsb.org/](https://search.rcsb.org/)
- RCSB Data API documentation: [https://data.rcsb.org/](https://data.rcsb.org/)
- Berman HM et al. The Protein Data Bank. *Nucleic Acids Res.* 2000;28(1):235-242. doi: [10.1093/nar/28.1.235](https://doi.org/10.1093/nar/28.1.235)

## Chat Usage

### Example Prompts

- "Download EGFR kinase structure 1M17 from PDB"
- "Search PDB for BRAF kinase structures with resolution below 2.0"
- "Find cryo-EM structures of the SARS-CoV-2 spike protein"

### Workflow Chaining

- **Feeds into**: DiffDock (docking), P2Rank (pocket prediction), PDB2PQR (electrostatics), MM-GBSA (rescoring)
- **Accepts from**: UniProt (cross-referenced PDB entries), BLAST (homologous structures)

## FAQ

### How is PDB Search different from ESMFold?

  PDB Search retrieves experimentally determined structures (X-ray, cryo-EM, NMR) from the RCSB database. ESMFold predicts structures from sequence using AI. Use PDB when an experimental structure exists; use ESMFold when no experimental structure is available.

### What resolution should I look for?

  Less than 2.0 Angstrom is high resolution (excellent for docking). 2.0-2.5 Angstrom is acceptable for most analyses. Above 3.0 Angstrom requires caution for detailed binding site analysis.

## Read next