Protein Bank Database

If you work with biomolecular structures, you need the Protein Data Bank (PDB). This guide explains what the Protein Data Bank is, how to search it, how to assess structure quality, and when to trust the data. It is written for life scientists, structural biologists, and bioinformatics beginners who want a practical, source bounded introduction. The PDB is the single global repository for experimentally determined three dimensional structures of proteins, nucleic acids, and complex assemblies. You will learn how to choose a structure, run a retrieval workflow, avoid common pitfalls, and understand the limits of structural interpretation.

At a Glance

For the most authoritative overview of the PDB and its management consult the NCBI Bookshelf entries on molecular biology databases NCBI Bookshelf.

Core Concepts

The Protein Data Bank (PDB) was established in 1971 and is now maintained by the Worldwide Protein Data Bank (wwPDB) consortium. Every entry receives a four character alphanumeric identifier called a PDB ID. For example, the structure of lysozyme is 1AKI. The database stores the three dimensional coordinates of atoms in a standard PDB file format or the newer mmCIF format. Each entry also contains the sequence of the macromolecule, the experimental method used, the resolution of the data, and the authors. This metadata is critical for evaluating the reliability of the structure.

The PDB is not a compilation of predicted models. It only houses experimental data. However, many predicted structures from AlphaFold and other methods are now deposited in a separate resource called the AlphaFold Protein Structure Database. The PDB itself focuses on experimentally validated structures. Researchers often use PDB structures as templates for homology modeling, docking studies, and structure based drug design. To get started with searching and analyzing PDB data the EMBL EBI training platform offers structured courses on structural bioinformatics EMBL-EBI Training. You should also become familiar with the validation reports that accompany each PDB entry. These reports highlight geometric outliers and poor density fitting.

Decision Points

When selecting a PDB structure you need to weigh several factors. The most important decision criterion is resolution. For X ray crystallography, resolution is reported in angstroms. A resolution of 2.0 A or better is generally considered high quality. For cryo EM the resolution is also reported but you must look at the map model correlation and the overall resolution estimate. For NMR structures, resolution is not applicable, instead, you evaluate the number of restraints and the ensemble RMSD.

Another decision point is the presence of ligands or cofactors. If you are investigating a protein ligand interaction you need a structure that contains the small molecule of interest. Check the chemical component dictionary to confirm the ligand identity. Decisions about which PDB entry to use also depend on whether the structure includes mutations. Many entries contain engineered mutations for crystallization. The original wild type sequence may differ. Use the sequence cross reference tools to check for mutations. The Galaxy Training Network provides practical tutorials on how to compare PDB sequences with UniProt entries Galaxy Training Network.

You must also decide between using a single structure or a set of structures from a family. For flexible proteins, multiple conformations may exist in the PDB. The best approach is to collect all relevant entries and perform a structural alignment to understand the conformational landscape.

Practical Workflow or Implementation Sequence

Follow this step by step workflow to retrieve and evaluate a PDB structure for your project.

1. Define your target. Write down the protein name, accession number, or PubMed citation. If you are studying a specific ligand, include its chemical name or SMILES.
2. Search the PDB. Go to RCSB PDB or PDBe and use the search bar. Use Boolean operators if needed. For example, search for "BRCA1 AND human AND resolution < 3.0". Filter by experimental method and resolution.
3. Review the summary page. Click on a candidate entry. Look at the resolution, R free, number of residues modeled, and any missing loops. Check the sequence alignment to your target.
4. Download the structure and validation report. Download the PDB file or mmCIF file. Also download the two page PDF validation report. This report contains key quality metrics such as clashscore, Ramachandran favored percentage, and MolProbity score.
5. Visualize and inspect. Use a program like PyMOL, ChimeraX, or VMD. Color by B factor to see which regions are mobile. Check the electron density if available. Many PDB entries provide a density map link.
6. Perform geometric validation. Use online tools such as MolProbity or the wwPDB validation server. Compare your own structure validation numbers against typical values for the method and resolution.
7. Document your selection. Write why you chose the structure, what quality metrics you used, and any known limitations. This step is critical for reproducibility.

For an in depth tutorial on implementing structural analysis workflows using Bioconductor packages for R visit the Bioconductor documentation Bioconductor. You can also find scripts for batch downloading and validation on the Galaxy Training Network.

Common Mistakes

Many new users make errors that compromise the reliability of their structural analysis. Here are the most frequent pitfalls and how to avoid them.

• Ignoring validation reports. Do not rely solely on resolution. A structure with low resolution but excellent geometry may be more reliable than a medium resolution structure with many outliers. Always check the clashscore and Ramachandran plot.
• Using the wrong chain or model. PDB files may contain multiple chains, asymmetric units, or several NMR models. Make sure you extract the relevant chain and the correct model (usually model 1 for NMR ensembles, but sometimes model 1 is the most representative).
• Overinterpreting side chain conformations. At resolutions below 3.0 A side chain positions are uncertain. Do not make mechanistic claims based on a single rotamer. Use multiple structures or biological validation.
• Confusing the PDB sequence with the wild type sequence. Because of crystallization tags, mutations, or engineered sequences, the sequence in the PDB file may not match the canonical protein. Always align with UniProt.
• Forgetting to check for biological assembly. The coordinates in the asymmetric unit may not represent the biologically relevant oligomer. Use tools like PISA to determine the correct assembly.
• Using an outdated PDB entry. Structures are sometimes superseded by newer, higher quality determinations. Check the "superseded by" field on the PDB page.

Another mistake is using PDB entries from the Sequence Read Archive (SRA) as a structural source. The SRA contains raw sequencing reads, not 3D structures. For structural data the PDB is the correct repository. For sequence data use the NCBI Sequence Read Archive NCBI Sequence Read Archive. The two resources serve different purposes. If you need both sequence and structure you must cross reference.

Limits and Uncertainty

No PDB structure is a perfect representation of a biological molecule. Every experimental method introduces uncertainty. X ray structures have crystal packing artifacts. NMR structures average over multiple conformations. Cryo EM maps have limited resolution for flexible regions.

You must also account for the limitations of validation metrics. A structure that passes all validation criteria can still be biologically irrelevant if the construct was truncated or mutated. The PDB does not verify biological function. It archives what the depositor provided. The validation reports are automated and may not detect systematic errors such as wrongly assigned atom types or ligand distortions.

The PDB is also incomplete. Many important proteins have not been solved. This is especially true for membrane proteins, large complexes, and disordered proteins. Use caution when extrapolating from a single structure to a whole protein family. The structural coverage in the PDB is biased toward easy to crystallize proteins. As a result, you should complement PDB data with sequence based predictions and biochemical experiments.

For example, the work on splice altering TP53 missense mutations shows that structure alone cannot predict functional impact. Structural analysis needs to be paired with transcriptomics and functional assays. That study used PDB entries but also relied on RNA sequencing data to identify cryptic splice sites Splice altering TP53 missense mutations as drivers of dysfunction and targets for RNA based therapy. Similarly, the integration of genetic evidence with structural data helps identify approved drug targets Integration of genetic evidence to identify approved drug targets. These examples remind us that the PDB is a tool, not a final answer.

Frequently Asked Questions

1. What is the difference between the PDB and the AlphaFold Database?
The PDB holds experimentally determined structures only. The AlphaFold Database holds computational predictions. Both can be useful, but predictions are not validated by experiment. For drug design, experimental PDB structures are usually preferred.

2. How do I find ligand binding sites in a PDB structure?
Use the PDBe or RCSB PDB website to view bound molecules. You can also run a tool like ProBis or fpocket. For a unified portal that integrates PDB and PubChem data see the Protein Data Bank Japan resource Protein Data Bank Japan: A unified portal for integrating structural and chemical data to explore protein ligand interactions in PDB and PubChem.

3. Can I use PDB structures for machine learning?
Yes, but you must curate the dataset carefully. Remove low resolution entries, filter by resolution and R free, and check for sequence redundancy. Tools like Scotty can help identify lattice coincidences and artifacts Scotty: lattice coincidences in the Protein Data Bank. For sequence level features, curated databases like FeSseqdb provide another layer FeSseqdb: a curated sequence level database and interpretable machine learning framework for identifying iron sulfur proteins.

4. What should I do if the PDB structure I need is missing?
You can try homology modeling using a template from the PDB. Use SWISS MODEL or MODELLER. Alternatively, search for a structure in the Electron Microscopy Data Bank (EMDB). If no template exists, you may need to solve the structure yourself or use computational predictions.

References and Further Reading

• NCBI Bookshelf. Molecular Biology Databases. NCBI Bookshelf
• EMBL EBI Training. Structural Bioinformatics. EMBL EBI Training
• Galaxy Training Network. Sequence Analysis and Structure Comparison. Galaxy Training Network
• Bioconductor. Structural Biology Packages. Bioconductor
• NCBI Sequence Read Archive. High throughput Sequencing Data. NCBI Sequence Read Archive
• Splice altering TP53 missense mutations as drivers of dysfunction and targets for RNA based therapy. Stem Cell Res Ther. 2025. PubMed
• Protein Data Bank Japan: A unified portal for integrating structural and chemical data to explore protein ligand interactions in PDB and PubChem. Protein Sci. 2025. PubMed
• Integration of genetic evidence to identify approved drug targets. Genome Med. 2025. PubMed
• Computational identification of putative GyrA and AcrR inhibitors from gut microbiota and natural products targeting multidrug resistant Salmonella enterica. Sci Rep. 2025. PubMed
• Scotty: lattice coincidences in the Protein Data Bank. Acta Crystallogr D Struct Biol. 2025. PubMed

Related Articles

• Protein Synthesis
• Incomplete Dominance Gene
• Cell Membrane Function Biology
• Dna Structure
• Protein Structure