Loading field map
Structural biology is the branch of biology that aims to determine the three-dimensional arrangement of atoms in biological macromolecules—primarily proteins, nucleic acids, and their complexes—and to use that information to explain how these molecules perform their functions. The field rests on a central premise: that a molecule’s shape, at atomic resolution, largely determines its chemical behavior, its interactions with partners, and its role in the living cell. Because most biological processes are carried out by molecular machines whose moving parts must fit together precisely, knowing the structure of these machines is often the key to understanding how they work, how they fail in disease, and how they might be manipulated by drugs.
Structural biology addresses a small set of enduring questions, though the tools used to answer them have changed dramatically. The first is simply descriptive: what does a given molecule look like? This includes the overall fold of a protein’s polypeptide chain, the positions of side chains, the location of bound metals or cofactors, and the arrangement of subunits in a multi-protein complex. The second question is mechanistic: how does the structure explain the molecule’s function? For an enzyme, this means identifying the active site and understanding how it stabilizes a transition state; for a motor protein, it means tracing how conformational changes convert chemical energy into mechanical work; for a receptor, it means seeing how ligand binding triggers a shape change that propagates across a membrane.
A third question concerns dynamics. Biological molecules are not static; they breathe, sample multiple conformations, and undergo large rearrangements during catalysis, signaling, or assembly. A single static structure is therefore a snapshot, and structural biologists increasingly ask how the ensemble of possible structures relates to function. A fourth question is evolutionary: how do structures change as sequences diverge, and how do conserved structural features constrain or enable new functions? Finally, there is an applied question: how can structural knowledge guide the design of drugs, engineered enzymes, or novel biomaterials? These questions are not pursued in isolation; a single study often moves from determining a structure to proposing a mechanism to suggesting a therapeutic strategy.
The field’s origins lie in the mid-twentieth century, when the first protein structures were solved. The key enabling technique was X-ray crystallography, which requires growing ordered crystals of a molecule and then illuminating them with X-rays. The pattern of diffracted X-rays, combined with knowledge of the molecule’s chemical composition, allows reconstruction of its electron density and hence its atomic positions. The first globular protein structures—myoglobin and hemoglobin—were determined in the late 1950s and early 1960s, revealing for the first time how a polypeptide chain folds into a compact, functional shape. These achievements established the core workflow of structural biology: purify a molecule, crystallize it, collect diffraction data, solve the phase problem, and build an atomic model.
For decades, crystallography was the dominant method, and its scope expanded steadily. Improvements in synchrotron X-ray sources, cryogenic sample cooling, and computational phasing methods made it possible to solve larger and more challenging structures. By the 1990s, structures of ribosomes—enormous molecular machines composed of dozens of proteins and several RNA molecules—were being determined, a feat that required heroic effort and marked the field’s coming of age. Alongside crystallography, nuclear magnetic resonance (NMR) spectroscopy emerged in the 1980s as a complementary method. NMR can determine structures of small to medium proteins in solution, and uniquely, it can report on dynamics and weak interactions. However, NMR is limited by molecular size and requires high protein concentrations and isotopic labeling.
A third major method, cryo-electron microscopy (cryo-EM), had a slower ascent. Early electron microscopy of biological samples suffered from radiation damage and poor contrast, limiting it to low-resolution shapes. The field was transformed in the 2010s by the development of direct electron detectors and improved image-processing algorithms. These advances made it possible to determine near-atomic-resolution structures of molecules that resisted crystallization, including large membrane proteins and flexible complexes. Cryo-EM does not require crystals; instead, samples are frozen in a thin layer of vitreous ice, and thousands of images of individual particles are averaged computationally to reconstruct a three-dimensional density map. This “resolution revolution” made cryo-EM a mainstream technique and, for many targets, the method of choice.
A fourth approach, computational structure prediction, has a longer history but became transformative only recently. For decades, predicting a protein’s three-dimensional structure from its amino acid sequence alone was considered a grand challenge. Early efforts used physics-based energy functions and comparative modeling when a homologous structure was known. The field advanced steadily but slowly. In the early 2020s, deep-learning methods, particularly AlphaFold, achieved remarkable accuracy in predicting structures for the vast majority of proteins in sequenced genomes. This development did not replace experimental structural biology—predicted models still lack the precision of experimental structures for many purposes, and they do not capture ligand binding, post-translational modifications, or conformational dynamics—but it fundamentally changed the landscape. Structure prediction now provides a starting point for nearly any protein of interest, and experimental methods are increasingly used to test, refine, and extend predictions.
The four approaches—crystallography, NMR, cryo-EM, and computation—are not rival schools in the sense of competing paradigms. They are complementary tools, each with distinct strengths and limitations, and modern structural biologists routinely combine them. The choice of method depends on the nature of the sample and the question being asked.
X-ray crystallography remains the gold standard for high-resolution structures of well-behaved, crystallizable molecules. Its strengths are atomic resolution, often better than 2 Å, and the ability to study very large complexes if they form ordered crystals. Its central limitation is the requirement for crystals, which can be difficult or impossible to obtain for membrane proteins, flexible complexes, or heterogeneous samples. Crystallography also reports a time- and space-averaged structure; conformational heterogeneity is largely hidden, though it can sometimes be modeled as alternate conformations or high B-factors.
NMR spectroscopy occupies a distinct niche. It is uniquely suited to studying proteins in solution under near-physiological conditions, and it can measure dynamics across a wide range of timescales, from picosecond bond vibrations to seconds-long conformational exchanges. NMR can also detect weak and transient interactions, such as those between a signaling protein and its many partners. Its limitations are size—practical upper limits are roughly 30–40 kDa for routine structure determination, though specialized techniques extend this—and the need for high concentrations of stable, isotopically labeled protein. NMR is therefore most powerful for small, soluble, dynamic proteins and for characterizing interactions and dynamics rather than for solving large complexes.
Cryo-EM has become the method of choice for large assemblies, membrane proteins, and samples that are too heterogeneous or too fragile to crystallize. It can handle molecular weights from a few hundred kilodaltons to megadalton complexes, and it can sort particles into different conformational states computationally, providing a movie-like view of a machine’s functional cycle. The resolution of cryo-EM has improved dramatically, with many structures now reaching 2–3 Å, but it generally remains slightly lower than the best crystallographic structures. Cryo-EM also requires substantial computational resources and expertise in image processing, and it is less well suited to very small proteins, though recent advances are pushing into that territory.
Computational methods span a wide range. Homology modeling, which builds a structure based on a related template, has been used for decades and remains valuable when a close relative is known. Physics-based molecular dynamics simulations provide a way to explore conformational dynamics and to compute free energies, complementing experimental structures with a time dimension. The recent deep-learning predictors, such as AlphaFold and related methods, generate structures from sequence with unprecedented accuracy for single chains and, increasingly, for complexes. These predictions are not experimental measurements; they are sophisticated inferences from evolutionary patterns and known structures. They are most reliable for well-folded, globular domains and less reliable for disordered regions, conformational flexibility, and the effects of mutations or ligands. The relationship between computation and experiment is now deeply collaborative: predictions guide experimental design, and experimental structures provide the training data and validation for predictors.
The current practice of structural biology is characterized by integration. A typical modern study might begin with a computational prediction, use cryo-EM to capture a large complex in multiple conformations, employ NMR or molecular dynamics to probe dynamics and interactions, and use crystallography to obtain a high-resolution view of a key active site. The boundaries between methods are porous, and many laboratories are expert in several.
The field has also expanded beyond its traditional focus on individual proteins. Structural biology now routinely addresses large macromolecular machines, such as the ribosome, the spliceosome, and the proteasome; membrane-embedded signaling complexes; and even intact cellular structures, such as nuclear pore complexes and cilia. It increasingly interfaces with cell biology, as cryo-electron tomography allows structures to be determined inside cells, albeit at lower resolution. It interfaces with genomics, as structure prediction is applied to entire proteomes. And it interfaces with drug discovery, where structures of drug targets bound to candidate compounds guide medicinal chemistry.
Several challenges remain. Intrinsically disordered proteins and regions, which lack a stable fold, resist traditional structure determination and require specialized methods. Membrane proteins, despite major advances, remain harder to study than soluble proteins. Conformational dynamics, especially large-scale rearrangements, are still difficult to capture comprehensively. And the gap between static structures and the complex, crowded environment of the living cell remains wide. Structural biology is therefore not a finished enterprise but a field whose tools are converging on a more complete, dynamic, and cellular view of molecular architecture. The ultimate goal is not merely a collection of structures but a mechanistic understanding of how the molecules of life work, fail, and can be repaired.