12.4 Integrating structural information from multiple techniques

X-ray Crystallography vs NMR Spectroscopy

X-ray crystallography and NMR spectroscopy are the two workhorses of structural biology, but each captures a different aspect of molecular reality. X-ray gives you a high-resolution snapshot of a molecule locked in a crystal lattice, while NMR shows you how that molecule behaves in solution, including its flexibility and motion. Neither technique alone tells the full story, which is why structural biologists routinely combine them with methods like cryo-EM and SAXS to build more complete and reliable models.

Strengths and Weaknesses of X-ray Crystallography

X-ray crystallography delivers atomic-level resolution (often better than 2 Å) for macromolecules like proteins and nucleic acids. It also scales well to very large complexes: the ribosome and entire virus capsids have been solved this way.

The main limitations:

• The molecule must form a well-ordered crystal, which can be extremely difficult for flexible proteins, membrane proteins, or intrinsically disordered regions.
• Crystal packing forces can distort the structure, so what you see may not perfectly reflect the molecule's native conformation.
• The resulting structure is essentially a time-averaged, static picture. You get very little information about dynamics or conformational flexibility.

Strengths and Weaknesses of NMR Spectroscopy

NMR determines structures in solution, which is much closer to physiological conditions. This means you can directly observe dynamics, conformational exchange, and flexibility on timescales from picoseconds to seconds.

NMR is particularly well-suited for:

• Studying smaller proteins and protein-ligand interactions (enzyme active sites, hormone-receptor binding)
• Measuring binding affinities, on/off rates, and conformational changes in real time
• Characterizing disordered or highly flexible regions that don't show up well in crystal structures

The trade-offs are significant, though. NMR is generally limited to proteins smaller than about 50 kDa (though specialized techniques like TROSY can push this higher). Resolution is lower than X-ray crystallography. And structure determination typically requires isotopically labeled samples ($^{13}\text{C}$, $^{15}\text{N}$), which adds cost and preparation time.

Complementary Structural Techniques

Combining Data from Multiple Techniques

No single method captures every aspect of a macromolecule's structure and behavior. Each technique occupies a different niche in terms of resolution, size range, and the type of information it provides:

A common strategy is to use X-ray or cryo-EM for the high-resolution framework, then layer in NMR data to understand which parts of the structure are flexible or undergo conformational changes. SAXS can verify that a crystal structure's shape matches what the molecule looks like in solution, catching cases where crystal packing has introduced artifacts.

Cryo-EM and SAXS are especially valuable for targets that resist crystallization, such as membrane proteins and large multi-subunit assemblies like viral capsids.

Integrating Data from Different Techniques

Combining data from different methods isn't as simple as overlaying two pictures. Each technique has its own assumptions, error sources, and resolution limits, so integration requires a deliberate workflow:

1. Collect data from each technique independently. Solve what you can at high resolution (X-ray or cryo-EM), and gather complementary data (NMR relaxation, SAXS profiles, cross-linking constraints).
2. Identify what each dataset contributes. For example, X-ray might give you the core fold, while NMR chemical shift perturbations reveal which surface residues contact a binding partner.
3. Use computational tools to merge the data into a unified model. Software like HADDOCK (for docking) or the Integrative Modeling Platform (IMP) can combine heterogeneous restraints into a single structural model.
4. Validate the integrated model against independent experiments. Mutagenesis data, biochemical assays, or cross-linking mass spectrometry results that were not used during model building serve as a reality check.

The validation step is critical. Without it, you risk building a model that fits your data but doesn't reflect the actual biology.

Computational Methods in Structural Biology

Role of Computational Methods in Data Integration and Refinement

Computational tools are the glue that holds integrative structural biology together. Raw experimental data from different techniques come in very different forms (electron density maps, NOE distance restraints, scattering curves), and software is needed to translate all of these into a coherent three-dimensional model.

Key computational approaches include:

• Molecular modeling and docking (Rosetta, Modeller, HADDOCK): These generate candidate structures consistent with experimental restraints. Rosetta, for instance, can model missing loops or predict how two proteins dock together using sparse experimental data as guides.
• Molecular dynamics (MD) simulations: MD refines static structures by simulating atomic motion over time. If NMR data show that a loop is flexible, MD can generate an ensemble of conformations that matches the experimental observations.
• Model quality assessment: Tools like MolProbity evaluate whether a structure makes physical sense by checking Ramachandran plot distributions (are backbone angles in allowed regions?), steric clash scores, and rotamer outliers. These checks catch errors that might slip through refinement.

Integrative Modeling Approaches

For large, complex systems where no single technique can solve the full structure, integrative modeling provides a formal framework for combining everything you know.

The Integrative Modeling Platform (IMP) is one widely used example. The general workflow:

1. Gather all available data: crystal structures of individual subunits, cryo-EM envelopes of the full complex, NMR-derived distances, SAXS profiles, cross-linking mass spectrometry distance constraints.
2. Translate each data type into spatial restraints: for instance, a cross-link between two residues means those residues must be within a certain distance.
3. Sample conformational space: the software explores millions of possible arrangements of the subunits, scoring each against all the restraints simultaneously.
4. Cluster and validate the results: the best-scoring models are grouped, and their consistency with data not used in modeling is checked.

This approach has been applied to systems that would be nearly impossible to solve by any single method, including the nuclear pore complex (a massive assembly of ~30 different proteins) and chromatin fiber organization.

Integrated Structural Analysis

Case Studies Demonstrating the Power of Integrated Structural Biology

Three landmark examples illustrate how integration works in practice:

The ribosome. Early crystal structures of individual ribosomal subunits (30S and 50S) provided atomic detail, but understanding the full 70S ribosome in different functional states required cryo-EM to capture conformational snapshots during translation. Biochemical crosslinking and footprinting data helped confirm subunit interfaces and tRNA binding sites. The result was a near-complete mechanistic picture of protein synthesis.

G protein-coupled receptors (GPCRs). These membrane proteins are targets for roughly a third of all approved drugs, yet they're notoriously hard to crystallize in active conformations. X-ray structures provided the first atomic views, but NMR spectroscopy revealed that GPCRs exist in dynamic equilibria between active and inactive states. Computational modeling then connected these structural snapshots into a continuous picture of receptor activation, directly informing drug design.

Intrinsically disordered proteins (IDPs). Proteins like alpha-synuclein (linked to Parkinson's disease) and the transactivation domain of p53 don't adopt a single folded structure, so X-ray crystallography largely fails. NMR spectroscopy characterizes the residue-level preferences within the disordered ensemble, SAXS provides the overall dimensions, and computational methods (ensemble generation and selection) combine these data into a representative set of conformations. This integrated approach revealed that "disordered" doesn't mean "random": these proteins often have transient structural preferences that are functionally important.

Importance of a Multidisciplinary Approach in Structural Biology

These case studies share a common lesson: the most impactful structural biology projects draw on multiple techniques and require collaboration across disciplines. A crystallographer, an NMR spectroscopist, a computational modeler, and a cell biologist working on the same problem will each contribute information that the others cannot access alone.

This multidisciplinary approach is increasingly the standard, not the exception. Funding agencies and journals now routinely expect structural claims to be supported by orthogonal methods. For students entering the field, fluency in more than one technique, and an understanding of how they complement each other, is becoming essential.