Determining protein structure

Size and mass

SDS-PAGE. Sodium dodecyl sulfate binds proteins at a roughly constant ratio, unfolding them and swamping their native charge, while a thiol reductant breaks disulfides. Migration through a polyacrylamide gel then depends on size: log(mass) is roughly linear in distance travelled, so marker proteins calibrate the gel. Isoelectric focusing separates instead by pI, and combining the two gives 2D gels.

Mass spectrometry. Proteins need soft ionisation:

• MALDI: the protein is co-crystallised with a UV-absorbing organic matrix; a laser pulse vaporises the matrix, carrying the protein into the gas phase, mostly as singly charged ions measured by time-of-flight.
• ESI: a solution is sprayed from a charged needle; as droplets evaporate, multiply protonated ions [M+zH]^{z+} form. A protein gives an envelope of peaks at different z. Two adjacent peaks at m/z values m_1 > m_2 with charges z and z+1 satisfy m_1 z - z = m_2(z+1) - (z+1), so z and M can be solved; software does this deconvolution routinely.

In peptide mass fingerprinting a protein is digested with trypsin (cleaving after Lys and Arg) and the fragment masses matched to a database; tandem MS fragments peptides along the backbone to read sequence and locate modifications.

Secondary structure: circular dichroism

Chiral molecules absorb left- and right-circularly polarised light differently. In the far UV (about 190–250 nm) the peptide amide chromophores in a regular array give characteristic CD spectra: an \alpha-helix shows two negative bands near 208 and 222 nm and a positive band near 192 nm; a \beta-sheet a single negative band near 217 nm; random coil a strong negative band near 200 nm. CD is fast, needs little protein and is ideal for following unfolding with temperature or denaturant, but it gives only overall percentages, not positions.

Atomic structures

X-ray crystallography. A single crystal diffracts X-rays into spots whose intensities are measured, but their phases are lost (the phase problem). Phases are recovered by molecular replacement with a related structure, by heavy-atom derivatives or by anomalous scattering (e.g. selenomethionine). An electron-density map is calculated, a model built into it and refined; resolution (in Å) and R-factors report quality.

NMR spectroscopy. For proteins up to a few tens of kDa, uniformly ^{15}N/^{13}C-labelled samples give multidimensional spectra. Assigned resonances provide NOE distance restraints (through-space, up to about 5 Å) and torsion-angle restraints from couplings and chemical shifts, from which an ensemble of structures is calculated. NMR works in solution and reports on dynamics.

Structural biology has shifted markedly in the last decade. Cryo-electron microscopy, driven by direct electron detectors and better image processing, now routinely reaches near-atomic resolution for large complexes and membrane proteins that resist crystallisation, and separates conformational states by classifying particle images. Crystallography remains the workhorse for high-throughput ligand screening, with fragment soaking campaigns at synchrotrons and serial femtosecond crystallography at X-ray free-electron lasers enabling time-resolved snapshots of reactions.

Machine-learning structure prediction (AlphaFold and successors, RoseTTAFold) gives accurate models for most single domains and many complexes, together with per-residue confidence estimates. Experimental methods increasingly validate, refine or extend these models: predicted structures solve phase problems by molecular replacement and fit into medium-resolution cryo-EM maps. Open questions include conformational ensembles, intrinsically disordered regions, ligand-induced changes and the effects of mutations, where predictions are much less reliable.

Integrative structural biology combines sparse data: cross-linking MS gives distance restraints, hydrogen–deuterium exchange MS maps solvent protection and dynamics, native MS weighs intact complexes and reports stoichiometry, and small-angle X-ray scattering gives overall shape. Solid-state NMR handles amyloid fibrils and membrane proteins in lipid bilayers, and solution NMR relaxation methods characterise sparsely populated, functionally important excited states.

Top-down proteomics, in which intact proteins are fragmented in high-resolution mass spectrometers, identifies proteoforms: the specific combinations of modifications present on individual molecules, which bottom-up peptide methods cannot link together.