Monomer, Dimer, or Something Messier? Predicting Quaternary State Before You Purify

A size-exclusion peak at the “dimer” volume does not prove a dimer. An elongated monomer can elute early, a fast monomer–dimer equilibrium can form one averaged peak, and a weakly associating tetramer can disappear on dilution. Quaternary state is a model built from multiple measurements—not a label read from one chromatogram.

Computational prediction is useful before purification because it exposes plausible assemblies, interfaces, and experiments. It is dangerous when a predicted oligomer is treated as established biology.

Key Takeaways

• Stoichiometry is usually an input to structure prediction, not an output. Run competing biological hypotheses.
• Sequence can suggest oligomerization but rarely settles it. Coiled coils, repeat symmetry, transmembrane helices, and conserved surface patches need structural context.
• Interface confidence must be separated from monomer confidence. High pLDDT can coexist with an invented assembly.
• Concentration and conditions matter. A protein may be monomeric at assay dilution and oligomeric during concentration or crystallization.
• Use orthogonal mass methods. SEC-MALS, native MS, AUC, or mass photometry answer questions that SEC retention alone cannot.
• Treat heterogeneity as information. Multiple species may reveal reversible equilibria, domain swapping, partial unfolding, or construct-induced association.

Quaternary State Is a Conditional Property

“This protein is a dimer” is incomplete without conditions. Association depends on:

• concentration;
• pH and ionic strength;
• ligand, cofactor, nucleotide, or substrate;
• redox state;
• membrane and lipid composition;
• temperature;
• post-translational modifications;
• construct boundaries and tags;
• competing partners.

Obligate oligomers remain associated across a broad condition range because subunit folding and assembly are coupled. Transient or regulatory oligomers may shift state across the exact concentration range used for purification and assays.

What the Sequence Can Tell You

Sequence features can generate hypotheses:

• long coiled-coil propensity suggests extended oligomeric interfaces;
• repeated heptad patterns suggest specific helix packing;
• signal peptides and transmembrane helices constrain orientation;
• paired cysteines may support covalent assemblies;
• conserved exposed hydrophobic patches may indicate interfaces;
• repeat proteins may favor symmetric oligomers;
• short linear motifs may mediate conditional interactions;
• low-complexity regions may drive concentration-dependent condensates or nonspecific association.

None is decisive alone. A hydrophobic patch may be a ligand site. A coiled coil may be intramolecular. A cysteine may remain reduced.

Run a Stoichiometry Sweep

For a suspected homooligomer, predict a biologically plausible set: monomer, dimer, trimer, tetramer, perhaps hexamer if the family supports it. For each hypothesis, compare:

• pTM and ipTM;
• inter-chain PAE;
• interface area and contact chemistry;
• symmetry;
• unsatisfied buried polar groups;
• conservation of interface residues;
• recurrence across seeds;
• compatibility with membrane or cellular geometry.

A rising confidence score with every added chain is not proof that the largest assembly is correct. More chains create more opportunities for plausible contacts. Ask whether the assembly closes coherently, preserves symmetry, and matches known mass or family biology.

Read the Interface, Not Just the Fold

AlphaFold-Multimer and related methods may fold each subunit well while remaining uncertain about assembly. For complexes, inspect the off-diagonal PAE blocks and interface-focused scores. Our guide to pTM, ipTM, and inter-chain PAE covers this in detail.

Characteristics of a more credible interface include:

• a recurring pose across seeds;
• a localized low inter-chain PAE region;
• sensible hydrophobic burial without exposed unsatisfied core chemistry;
• complementary charge and hydrogen-bond networks;
• conserved interface residues;
• no conflict with known active sites, ligands, membranes, or partners;
• agreement with homologous assemblies.

Warning signs include:

• different top models use unrelated surfaces;
• only terminal tags or disordered tails make contact;
• the interface is tiny relative to the assembly;
• membrane proteins meet on impossible sides;
• a predicted oligomer blocks known function;
• one chain unfolds to create the contact.

Experimental Methods Answer Different Questions

Size-exclusion chromatography

SEC reports hydrodynamic size. Shape, glycosylation, disorder, and column interaction affect retention. Use it to separate species and assess homogeneity, not to assign exact molecular mass alone.

SEC-MALS

Multi-angle light scattering estimates molar mass across an elution peak. It is powerful for distinguishing compact and elongated species, provided concentration and refractive-index inputs are appropriate.

Analytical ultracentrifugation

AUC can resolve distributions and association equilibria over a concentration range. It is especially useful when species interconvert.

Native mass spectrometry

Native MS can reveal stoichiometries and bound ligands with high mass resolution, though ionization and gas-phase behavior require careful interpretation.

Mass photometry

Mass photometry measures individual particles near surfaces at low concentrations. It is rapid and sample-efficient, with practical lower-mass and concentration limits.

Cross-linking and structural methods

Cross-linking MS constrains interfaces; SAXS reports global shape; cryo-EM and crystallography can show assemblies directly. Crystal packing still requires distinguishing biological interfaces from contacts, the problem addressed by tools such as PISA (Krissinel and Henrick, 2007).

Design the Purification to Preserve the Question

If you want to know the native quaternary state, avoid accidentally forcing an answer.

• Record concentration at every step.
• Compare samples before and after concentration.
• Minimize high local concentrations on affinity resin.
• Test tag-cleaved and tagged material when the tag could oligomerize.
• Preserve required cofactors or ligands.
• Control redox conditions for cysteine-rich proteins.
• Use physiologically relevant salt where possible.
• Check whether dilution shifts the state.

For membrane proteins, detergent micelles inflate apparent size and complicate SEC interpretation. Use appropriate standards or orthogonal mass measurements, and consider nanodisc or native-lipid contexts.

A Decision Table for Common Observations

Worked Example: The “Dimeric” Enzyme

A 55 kDa enzyme elutes near a 100 kDa globular standard. AlphaFold-Multimer predicts a plausible dimer with ipTM 0.69, but the top five models split between two interfaces.

The team should not start interface mutagenesis yet.

1. SEC-MALS measures 61 kDa across the main peak: consistent with an elongated monomer.
2. SAXS supports an extended two-domain shape.
3. The dimer predictions use a flexible terminal domain with high inter-chain PAE.
4. Removing the terminus raises apparent multimer confidence but creates a construct not present in the measurement.
5. Mass photometry at assay concentration also shows predominantly monomer.

The correct conclusion is not “AlphaFold failed.” It is that the sequence can support geometrically plausible contacts, but the solution evidence favors a monomer under the tested conditions.

Now reverse the outcome: if SEC-MALS gives 108 kDa, native MS shows a dimer, and one predicted interface recurs across seeds, the same model becomes a strong basis for targeted mutagenesis.

When the Answer Is “Something Messier”

Heterogeneity can arise from:

• reversible monomer–oligomer equilibrium;
• multiple biological stoichiometries;
• domain swapping;
• partial proteolysis;
• disulfide scrambling;
• nucleic-acid bridging;
• aggregation after partial unfolding;
• mixed post-translational states;
• a tag or linker creating an artificial contact.

Do not force one structural story onto a broad peak. First determine whether species interconvert and whether activity tracks one state.

Integrating Prediction in Orbion

Orbion’s AstraSUIT can surface quaternary-state hypotheses from sequence context. Those hypotheses can then be compared with AlphaFold-Multimer, inter-chain PAE, interface contacts, and buried surface area. The result should be a ranked set of assemblies with explicit reasons—not a single unqualified oligomer label.

The best pre-purification output is an experimental plan: which concentrations to measure, which stoichiometries to model, which tags to control, and which interface changes would discriminate between assemblies without unfolding the monomer.

Bottom Line

Predicting quaternary state before purification is valuable because it makes your experiments sharper. Sequence and structure models generate plausible assemblies. Inter-chain confidence and interface chemistry rank them. Orthogonal mass and shape measurements decide which survive.

Never let SEC retention or one AlphaFold-Multimer run settle the question alone. Make competing stoichiometries explicit, test the state across relevant conditions, and treat disagreement as a clue to the biology.

References

1. Evans R, et al. Protein complex prediction with AlphaFold-Multimer. bioRxiv. 2022. doi:10.1101/2021.10.04.463034
2. Bryant P, Pozzati G, Elofsson A. Improved prediction of protein-protein interactions using AlphaFold2. Nature Communications. 2022. doi:10.1038/s41467-022-28865-w
3. Krissinel E, Henrick K. Inference of macromolecular assemblies from crystalline state. Journal of Molecular Biology. 2007. doi:10.1016/j.jmb.2007.05.022
4. Schuck P. Size-distribution analysis of macromolecules by sedimentation velocity ultracentrifugation and Lamm equation modeling. Biophysical Journal. 2000. doi:10.1016/S0006-3495(00)76713-0
5. Young G, et al. Quantitative mass imaging of single biological macromolecules. Science. 2018. doi:10.1126/science.aar5839