Exploring structures

Why it matters

More than 200,000 experimentally determined structures of proteins, nucleic acids and their complexes are freely available in the Protein Data Bank (PDB), plus over 200 million predicted structures in the AlphaFold database. Before starting an experiment, a biochemist checks whether a structure of the protein (or a relative) exists, where the active site is, which residues contact a ligand, and how reliable the model is. Reading a structure entry and moving around it in a viewer is an everyday research skill.

The PDB and its entries

The PDB is run by the worldwide PDB (wwPDB) partnership; the US portal is RCSB PDB at https://www.rcsb.org/ (Checked on 2026-10-23.). Each entry has a four-character PDB ID, such as 4HHB (human deoxyhemoglobin), and a summary page listing:

• Macromolecules: each chain (A, B, C...) and what it is.
• Ligands: small molecules, ions and cofactors (heme, ATP, metal ions), each with a three-letter code.
• Experimental method: X-ray crystallography, cryo-EM or NMR.
• Resolution (for X-ray and cryo-EM) in ångströms (Å). Smaller is better: around 1.5 Å shows individual atoms clearly; around 3.5 Å shows the chain path and large side chains, with less certainty.
• Validation report: global quality indicators (for X-ray, R-work and R-free; Ramachandran outliers; clashscore) compared with other entries.
• The primary citation for the paper that reported it.

Coordinates download as mmCIF (the current standard) or the legacy PDB format: plain-text files listing each atom's position, occupancy and B-factor (how much the atom's position is smeared, often by motion).

Experimental versus predicted

AlphaFold DB (https://alphafold.ebi.ac.uk/, Checked on 2026-10-23.) provides predicted models. Each residue has a pLDDT confidence score from 0 to 100: above 90 is very high confidence, below 50 often means a disordered region. Predicted models do not include ligands or show alternative conformations, and they are hypotheses to test, not measurements.

Viewers

• Mol* (Mol-star) runs in the browser on RCSB and PDBe pages. No install.
• PyMOL and ChimeraX are desktop programs widely used for figures and analysis. ChimeraX is free for non-commercial use; PyMOL has an open-source version and a licensed version.

Common viewer actions:

• Change representation: cartoon (shows secondary structure), sticks or ball-and-stick (atoms), surface (shape and pockets).
• Colour by chain, secondary structure, element, hydrophobicity or B-factor.
• Select residues or ligands and show what lies within a few ångströms.
• Measure distances and angles between atoms.

Worked example: hemoglobin, 4HHB

On the RCSB page for 4HHB you will find:

• Two kinds of chain: alpha (chains A and C) and beta (chains B and D), so the quaternary structure is a heterotetramer.
• Ligand HEM (heme), one per chain, each with an iron atom.
• Method: X-ray diffraction, at a resolution of about 1.7 Å.

In the 3D viewer, colour by chain to see the tetramer. Focus on a heme: the iron sits in the flat porphyrin ring, and one histidine (the proximal histidine) coordinates the iron from one side. Because 4HHB is the deoxy form, no oxygen is bound on the other side. Measuring the iron to histidine nitrogen distance gives roughly 2 Å, a typical coordination bond length. You have just related a structure to its function: this is the site where oxygen binds.

Reading quality critically

• Check resolution and the validation sliders before trusting fine details.
• Missing residues (gaps in the chain) usually mean flexible, disordered regions.
• High B-factors flag mobile or uncertain parts.
• A ligand's position in a low-resolution map can be ambiguous; look at the density fit if the viewer offers it.
• A crystal structure is one snapshot, possibly influenced by crystal packing or by the construct used (truncations, mutations, tags).

Common mistakes

• Treating a predicted model as experimental. Check the source and pLDDT.
• Ignoring the biological assembly. The file may hold one copy (asymmetric unit) or several; RCSB shows the biologically relevant assembly separately.
• Over-reading low resolution side-chain positions or water molecules.

Recap

• Use RCSB PDB entries for method, resolution, chains, ligands and validation.
• AlphaFold models come with per-residue confidence (pLDDT).
• In a viewer: change representation, colour, select and measure, and always check quality before drawing conclusions.

Try it

On the Do tab, explore 4HHB in the free Mol* viewer in your web browser.