- The three-dimensional structure of a protein provides information that cannot always be obtained from its amino acid sequence alone. Once a protein structure has been determined experimentally or predicted computationally, it becomes possible to investigate its domains, active sites, binding pockets, interaction surfaces, conformational states, and structural relationships with other proteins. Protein structure databases provide the resources needed to store, organize, search, visualize, compare, and interpret this structural information. They form an important connection between experimental structural biology, protein sequence analysis, computational modeling, and bioinformatics.
- The most fundamental resource in structural biology is the Protein Data Bank (PDB). The PDB is a worldwide repository of experimentally determined three-dimensional structures of biological macromolecules. These structures have been obtained using techniques such as X-ray crystallography, nuclear magnetic resonance spectroscopy, and cryo-electron microscopy. Each deposited structure is assigned a unique identifier, allowing researchers to retrieve the corresponding structural data and investigate the molecule in detail.
- A PDB entry can contain considerably more information than simply the coordinates of a protein. Depending on the structure, an entry may include the amino acid sequence, three-dimensional atomic coordinates, experimental method, resolution or other quality information, ligands, nucleic acids, ions, biological assemblies, experimental conditions, and information about the molecules included in the structure. This makes structural databases valuable sources of both structural and biological information.
- Protein structures are commonly represented as atomic coordinates. Each atom is assigned a position in three-dimensional space, allowing software to reconstruct the molecular structure. These coordinates can be visualized using molecular graphics programs, where proteins can be displayed as cartoons, surfaces, sticks, spheres, or other representations. Different visualization styles emphasize different structural features. A cartoon representation can highlight alpha helices and beta sheets, while a surface representation can reveal pockets, cavities, and molecular interfaces.
- The PDB contains structures from a wide range of biological systems, including individual proteins, protein complexes, nucleic acids, membrane proteins, enzymes, receptors, antibodies, molecular machines, and protein-ligand complexes. Consequently, it is not simply a database of isolated protein structures. It represents a large collection of experimentally characterized molecular assemblies and provides an important foundation for structural and molecular biology.
- An important distinction must be made between experimental structures and predicted structures. Experimental structures are derived from measurements using structural biology techniques, whereas predicted structures are generated computationally. Predicted models may be extremely useful, particularly for proteins without experimental structures, but they should be interpreted according to their prediction confidence and the biological question being investigated.
- The growth of AI-based protein structure prediction has greatly expanded the amount of structural information available to researchers. Resources associated with predicted structures can provide models for proteins that have not yet been experimentally characterized. These models can be particularly valuable when combined with sequence-based annotations, protein family information, domain predictions, and evolutionary conservation.
- A protein structure database can therefore be viewed as one part of a broader structural bioinformatics ecosystem. Researchers may begin with a protein sequence, search for experimentally determined structures, identify related proteins, examine predicted models, compare structural folds, visualize conserved residues, and investigate functional sites. Different resources support different stages of this process.
- Structural databases are closely connected to sequence databases. A structure is usually associated with one or more biological sequences, and researchers often move between sequence and structural information during analysis. A protein sequence can be used to search for structurally characterized homologues, while a known structure can reveal the sequence and domain organization of the corresponding protein. This relationship allows sequence-based and structure-based methods to complement one another.
- One important application is structure-based protein identification. A researcher may have an unknown protein sequence and discover that it resembles a protein whose structure has already been deposited in the PDB. If the evolutionary relationship is sufficiently strong, the known structure may provide a useful template for homology modeling. Even when sequence similarity is limited, structural similarity can sometimes reveal a common evolutionary fold.
- Structural classification systems provide another layer of information. Proteins can be grouped according to their structural features, including their overall folds, domain organization, and evolutionary relationships. Systems such as SCOP and CATH have played important roles in organizing protein structures into hierarchical classifications. These classifications help researchers understand how apparently different proteins can share structural architectures and how protein folds are related through evolution.
- Structural classification is particularly useful because proteins with similar structures do not always have highly similar sequences. Evolutionary changes can alter many amino acid positions while preserving the overall three-dimensional fold. Consequently, structural classification can reveal relationships that are difficult to detect using simple sequence comparison.
- The concept of a protein fold is central to structural classification. A fold describes a recurring three-dimensional arrangement of secondary-structure elements such as alpha helices and beta sheets. Different proteins can sometimes adopt similar folds even when their sequences and functions have diverged substantially. Studying recurring folds provides insight into protein evolution and the structural constraints that shape protein families.
- Structural comparison is another major application of protein structure databases. Two proteins can be aligned in three-dimensional space to determine whether their structures are similar. This is called structural alignment. Unlike sequence alignment, which compares the order of amino acids, structural alignment compares the spatial arrangement of residues or structural elements.
- Structural alignment can be particularly valuable when sequence similarity is weak. If two proteins have substantially different sequences but retain a similar arrangement of secondary-structure elements and conserved residues, structural comparison may reveal an evolutionary relationship that is not obvious from sequence analysis. Structural alignment can therefore complement BLAST searches, multiple sequence alignments, and Profile Hidden Markov Model analysis.
- The comparison of protein structures can also reveal differences that are biologically important. Two homologous proteins may share a common overall fold while differing in the shape of an active site, ligand-binding pocket, or interaction surface. These local structural differences can contribute to differences in substrate specificity, molecular recognition, regulation, or biological function.
- Protein structure databases are also important for studying protein domains. A multidomain protein may contain several regions that correspond to distinct structural units. Structural resources can help researchers determine whether predicted domains adopt known folds, how their boundaries are defined, and how the domains are arranged within the complete protein. This provides structural evidence that complements domain predictions from resources such as Pfam, InterPro, SMART, and CDD.
- The relationship between motifs and three-dimensional structure can also be investigated using structural databases. A conserved motif identified from sequence analysis can be mapped onto a known structure to determine where the corresponding residues are located. The residues may form part of an active site, ligand-binding pocket, structural core, or protein interaction surface. This provides a way to connect a short linear sequence pattern with its actual molecular role.
- Structural databases are especially valuable for studying enzymes. An experimentally determined enzyme structure can reveal the arrangement of catalytic residues, substrate-binding regions, cofactors, metal ions, and other structural components required for activity. When structures of related enzymes are available, researchers can compare their active sites to investigate differences in substrate specificity and catalytic mechanisms.
- The same approach can be applied to receptors and signaling proteins. Structural information can reveal ligand-binding pockets, interaction interfaces, conformational states, and regulatory regions. For membrane receptors, structures can provide information about transmembrane helices and extracellular or intracellular domains. Such information can be particularly valuable for understanding signal transduction.
- Protein-ligand structures provide another important category of structural information. A structure may show a protein bound to a small molecule, peptide, nucleotide, metal ion, substrate, inhibitor, or other ligand. These structures can reveal the precise molecular environment in which binding occurs. Comparing ligand-bound and ligand-free structures can also provide information about conformational changes associated with molecular recognition.
- Structural databases therefore have a major role in drug discovery. Researchers can inspect experimentally determined structures of drug targets, identify binding pockets, compare ligand interactions, and investigate how structural differences influence binding. When no experimental structure is available, predicted models may provide preliminary structural hypotheses that can guide computational investigations.
- Protein structures are also increasingly used in genetic variant interpretation. A sequence variant can be mapped onto a three-dimensional structure to determine whether it occurs within a conserved domain, active site, ligand-binding pocket, protein interface, transmembrane segment, or structural core. Comparing the position of a variant across related protein structures can provide additional evolutionary and structural context.
- However, structural location should not be interpreted in isolation. A variant located within a protein domain or active site may be biologically important, but its clinical significance requires additional evidence. Structural information is therefore best viewed as one component of a broader variant-interpretation framework that can include genetic, population, evolutionary, functional, and clinical evidence.
- Structural databases also support comparative genomics. Proteins from different organisms can be compared at the structural level to determine which features have been conserved during evolution. Conserved folds, domains, active sites, and interaction surfaces can reveal functional constraints that extend across species.
- Another important use is the investigation of protein-protein interactions. Structures of protein complexes can reveal the interfaces through which proteins interact. Researchers can identify contact residues, hydrogen bonds, hydrophobic interactions, electrostatic interactions, and other structural features that stabilize the complex. Mutations affecting these interfaces can then be investigated using structural and experimental approaches.
- The interpretation of structural data requires attention to experimental quality. Not every region of an experimental structure is necessarily resolved equally well. Flexible loops, terminal regions, disordered segments, and mobile domains may be poorly defined or absent from an experimentally determined structure. Resolution and other experimental metrics therefore provide important information about how confidently particular structural features can be interpreted.
- The same principle applies to predicted structures. A predicted model may contain high-confidence regions corresponding to stable domains and lower-confidence regions corresponding to flexible or disordered segments. Researchers should examine the confidence information associated with predicted models rather than treating the entire structure as equally reliable.
- Structural visualization is an essential part of this process. Molecular visualization software allows researchers to rotate structures, zoom into active sites, measure distances, examine interactions, compare conformations, and map sequence features onto three-dimensional models. Visualization can transform an abstract list of conserved residues into a spatial understanding of how those residues work together.
- A useful structural-analysis workflow therefore begins with a protein sequence and proceeds through several complementary resources. Sequence similarity searches can identify homologues. Multiple sequence alignment and Profile HMMs can reveal evolutionary conservation. Protein domain databases can identify conserved domains. Motif and signature analysis can highlight important sequence features. Structural databases can then be searched for experimentally determined structures or related folds. Predicted structures can provide additional information when experimental structures are unavailable. Finally, visualization and structural comparison can connect all of these observations.
- This workflow illustrates why structural bioinformatics should not be considered separate from sequence bioinformatics. Protein sequence and structure represent different levels of information about the same biological molecule. Sequence analysis is particularly powerful for identifying evolutionary relationships and conserved residues, while structural analysis reveals their spatial organization and molecular context.
- The increasing availability of structural models is also changing the scale of structural biology. Instead of studying only a relatively small number of experimentally characterized proteins, researchers can now investigate structural hypotheses across entire protein families and proteomes. This creates opportunities for large-scale analysis of protein evolution, functional annotation, genetic variation, molecular interactions, and disease mechanisms.
- At the same time, the distinction between known structure, homology model, and AI-predicted structure remains important. A structure experimentally determined for the exact protein provides one type of evidence. A homology model based on a related experimental structure provides another. An AI-generated prediction provides a computational estimate based on learned sequence and structural relationships. These sources can complement one another, but they should not be treated as interchangeable.
- Protein structure databases therefore act as an information bridge connecting experimental structural biology with modern computational biology. They allow researchers to move from amino acid sequences to three-dimensional structures, from structures to functional sites, and from individual proteins to evolutionary families and molecular systems. They also provide the reference data needed to develop and evaluate new structure prediction methods.
- The broader progression of the protein-analysis series can now be viewed as a connected information hierarchy. Protein sequences provide the starting point for identifying homologues and evolutionary relationships. Sequence alignment reveals similarity and conservation. Protein families organize related sequences, while protein domains identify recurring structural and functional units. Motifs and signatures identify smaller conserved features. Domain architecture describes how these elements are organized within a protein. Protein structure prediction and homology modeling provide three-dimensional models, while protein structure databases and structural bioinformatics allow these models and experimental structures to be searched, compared, visualized, and interpreted.
