- A drug rarely acts in isolation on a single molecular target. Although drug discovery often begins by identifying a protein that plays an important role in a disease process, the biological effects of a drug ultimately depend on a network of interactions involving proteins, enzymes, receptors, ion channels, transporters and other biomolecules. Understanding these drug-target interactions is therefore essential for connecting molecular binding with cellular responses, therapeutic effects, adverse effects and drug resistance. This broader perspective is known as polypharmacology, which describes the ability of a drug or chemical compound to interact with multiple biological targets.
- The molecular interaction between a drug and a protein can be considered at several levels. At the structural level, a ligand may occupy a binding pocket and form hydrogen bonds, hydrophobic contacts, electrostatic interactions, van der Waals interactions, aromatic interactions or metal coordination with specific residues. At the thermodynamic level, binding free-energy calculations can help estimate the favorability of this interaction. At the cellular level, however, the consequences of binding depend on protein abundance, localization, pathway connectivity, downstream signaling and the concentration of the drug. Thus, a strong biochemical interaction does not automatically imply a strong biological effect.
- A drug target is generally a biological molecule whose modulation contributes to a pharmacological effect. Proteins constitute the majority of classical drug targets, including enzymes, G protein-coupled receptors, nuclear receptors, ion channels, transporters and signaling proteins. Some therapeutic compounds can also interact with nucleic acids, lipids or other cellular components. The concept of a target is therefore functional as well as structural: a protein may bind a drug, but whether that interaction is biologically relevant depends on its role in the organism and the cellular context.
- Drug-target interactions can involve different types of molecular mechanisms. A drug may inhibit an enzyme, activate or block a receptor, modulate an ion channel, alter the activity of a signaling protein or interfere with a protein-protein interaction. Some compounds bind to the active site, directly competing with a substrate or cofactor. Others bind to an allosteric site, causing a conformational change that modifies activity at a different region of the protein. Some drugs form reversible interactions, whereas others can form covalent bonds with specific residues.
- The structural concepts developed earlier in this series are directly relevant to drug-target interactions. Protein domains, protein motifs and protein domain architecture influence the organization of functional regions within a target. Conserved residues can contribute to catalytic activity or ligand recognition, while variable residues surrounding a binding pocket can determine selectivity between closely related proteins. Structural alignment can reveal similarities and differences between protein targets, and protein structure visualization can help examine how ligands interact with individual residues.
- The process of identifying and characterizing a drug-target interaction often begins with the protein sequence. Protein sequence analysis, protein-family classification and domain annotation can reveal the evolutionary identity of a potential target. Multiple sequence alignment can identify conserved residues, while Profile Hidden Markov Models can help recognize protein families and domains. Structural information can then reveal the three-dimensional organization of the target and identify potential binding pockets.
- When an experimentally determined structure is available, the Protein Data Bank provides a valuable source of structural information. Structures containing bound ligands can reveal experimentally observed binding modes and interactions. When an experimental structure is unavailable, homology modeling or AI-based protein structure prediction can provide structural hypotheses. These models can subsequently be examined using molecular docking, molecular dynamics and other computational approaches.
- Molecular docking can be used to investigate how a compound may interact with a protein target. Docking generates possible ligand poses and uses scoring functions to prioritize them. When multiple proteins are considered, docking can also provide an initial hypothesis about potential off-target interactions. However, docking scores alone are insufficient to establish whether an interaction occurs biologically. Experimental binding measurements and functional assays are needed to confirm computational predictions.
- Molecular dynamics simulations add another dimension by examining how protein-ligand complexes behave over time. A ligand may remain stably positioned within a pocket, undergo substantial rearrangement or occasionally leave the binding site. Protein side chains, loops and domains can also change conformation. These dynamic effects can influence the stability and accessibility of binding sites and may explain why a ligand interacts differently with related proteins.
- Binding free-energy calculations provide a further quantitative layer. Methods such as MM-GBSA, MM-PBSA, free-energy perturbation and thermodynamic integration can be used to estimate relative or absolute binding energetics under appropriate conditions. Comparing predicted binding energetics across several protein targets can help investigate potential selectivity or off-target interactions. Nevertheless, computational affinity predictions remain estimates and must be interpreted alongside experimental evidence.
- The distinction between target affinity and target engagement is particularly important. A drug may show measurable biochemical affinity for a protein in an isolated assay, but the protein may be expressed at low levels in the relevant tissue or located in a cellular compartment that the drug cannot efficiently reach. Conversely, a target with moderate biochemical affinity may produce substantial biological effects if the target is highly abundant, readily accessible and strongly connected to the relevant signaling pathway.
- Drug concentration is another important factor. A compound may be highly selective at its therapeutic concentration but interact with additional proteins when its concentration increases. Consequently, a drug’s apparent selectivity depends not only on its intrinsic binding properties but also on exposure, distribution, metabolism and the concentrations reached in different tissues.
- This leads to the concept of a selectivity profile. Instead of asking only whether a drug binds its intended target, researchers can ask how its interaction strength varies across a panel of related and unrelated proteins. A kinase inhibitor, for example, may be designed to target one kinase but interact with other members of the kinase family because their ATP-binding sites share conserved structural features. Understanding these interactions can help explain both therapeutic activity and adverse effects.
- Off-target interactions occur when a compound interacts with biological molecules other than the intended target. Off-target binding is not necessarily harmful. In some cases, an additional interaction may contribute to the desired therapeutic effect. In other cases, it can produce unwanted pharmacological effects or toxicity. Identifying these interactions is therefore an important part of drug development.
- Polypharmacology provides a broader framework for understanding these effects. Rather than treating a drug as a molecule that acts on one protein, polypharmacology considers a drug as a perturbation of a network of molecular targets. The biological response can emerge from the combined effects of several interactions. This is particularly relevant for diseases involving multiple pathways, complex signaling networks or heterogeneous molecular mechanisms.
- Interestingly, polypharmacology is not always an undesirable property. Some therapeutic strategies intentionally seek multi-target activity. A single compound that modulates several disease-relevant proteins may influence multiple components of a pathological pathway. In such cases, the objective is not necessarily to maximize target selectivity but to achieve an appropriate and controlled pattern of target engagement.
- The distinction between selectivity and promiscuity is therefore context-dependent. A compound that interacts with many targets may be problematic if those interactions are unrelated to the therapeutic mechanism. However, multiple carefully characterized interactions may be useful when they contribute to a desired biological effect. The important question is not simply how many proteins a compound binds, but which proteins it interacts with, at what concentrations, and with what functional consequences.
- Protein families are particularly important when analyzing drug selectivity. Related proteins often share conserved structural features because they evolved from common ancestors. These conserved regions can create similar ligand-binding environments. At the same time, less-conserved residues around the binding pocket can provide opportunities for selective recognition. Structural comparison can therefore identify regions where closely related targets differ.
- Binding-pocket analysis is useful in this context. Two proteins may have highly similar overall structures but different pocket volumes, shapes, charge distributions or flexible loops. A ligand that fits one pocket may therefore interact less favorably with another. Structural alignment combined with sequence conservation and molecular visualization can help identify the molecular basis of this selectivity.
- The same principle applies to protein domain architecture. Two proteins may share one catalytic domain but differ substantially in regulatory or interaction domains. A drug that binds the conserved catalytic domain may therefore affect both proteins, whereas a molecule designed to recognize a unique regulatory domain could achieve greater selectivity. Domain organization can consequently provide important information for drug-target design.
- Allosteric sites offer another strategy for achieving selectivity. Active sites are often strongly conserved among related proteins because they perform essential biochemical functions. Allosteric regions can be more variable. A compound that binds an allosteric pocket may therefore distinguish between closely related proteins more effectively than a compound targeting a highly conserved active site.
- Drug-target interactions can also involve protein-protein interaction interfaces. Many biological processes depend on proteins binding to one another, and disrupting or stabilizing these interactions can alter signaling pathways. Because protein-protein interfaces are often broad and relatively shallow compared with conventional enzyme pockets, designing small molecules against them can be challenging. Structural biology, molecular docking, molecular dynamics and fragment-based approaches can nevertheless help identify suitable binding regions.
- Drug interactions with ion channels and transporters present additional complexities. These proteins can contain transmembrane domains and undergo substantial conformational changes during their functional cycles. A drug may bind preferentially to one conformational state and thereby alter channel opening, closing or transport. Molecular dynamics simulations can be particularly informative for examining these dynamic systems.
- G protein-coupled receptors (GPCRs) provide another important example of conformationally dynamic drug targets. Different ligands can stabilize different receptor conformations, which may preferentially activate particular downstream signaling pathways. This phenomenon is sometimes described as biased signaling or ligand-directed signaling. Structural studies of receptor-ligand complexes can help explain how chemically different ligands influence receptor conformation and signaling.
- Enzymes provide a more straightforward example of drug-target relationships. An inhibitor may compete with the natural substrate, bind to a regulatory site or form a covalent interaction with an essential residue. Sequence conservation can identify catalytic residues, while structural analysis can reveal how the inhibitor occupies the active site. Molecular docking and molecular dynamics can then help investigate alternative binding modes and conformational effects.
- Drug-target analysis is also important for drug resistance. Genetic mutations can modify a drug-binding pocket, alter protein conformation or change the interaction network surrounding a target. If a mutation reduces drug binding while preserving sufficient protein function, cells carrying that mutation may become less sensitive to treatment. Structural analysis of wild-type and mutant proteins can help explain these changes, while free-energy calculations can provide estimates of how specific substitutions may influence binding.
- The same framework can be applied to naturally occurring genetic variants. A variant may alter an amino acid directly involved in drug binding, indirectly change the structure of a binding pocket or modify protein dynamics. Integrating sequence conservation, protein structure, molecular dynamics and experimental pharmacology can help investigate how such variants influence drug response. Computational predictions should again be regarded as mechanistic hypotheses rather than definitive clinical conclusions.
- Drug-target relationships can be represented computationally using a drug-target interaction network. In such a network, drugs and biological targets are represented as nodes, while experimentally observed or computationally predicted interactions are represented as edges. Additional information can be incorporated, such as binding affinity, target expression, pathway membership, tissue distribution or functional consequences.
- Network representations make it possible to connect molecular interactions with biological pathways. A drug may bind several proteins, and each protein may participate in several signaling or metabolic pathways. The resulting network can help researchers investigate mechanisms of action, possible off-target effects, drug combinations and pathway-level responses.
- Chemical similarity can also provide clues about drug-target relationships. Structurally related compounds often interact with overlapping sets of proteins, although small chemical modifications can substantially change affinity or selectivity. Chemical fingerprints, molecular descriptors and machine-learning methods can therefore be used to identify compounds that may share target profiles.
- Conversely, known ligands can be used to predict potential targets for a new compound. This is the basis of ligand-based target prediction. If a new molecule resembles compounds known to bind a particular protein family, computational models may prioritize that family for experimental testing. Such predictions are especially useful when the three-dimensional structure of the target is unavailable.
- Structure-based target prediction provides a complementary strategy. Instead of starting from ligand similarity, researchers can examine protein structures and identify pockets that could accommodate a compound. Docking, pharmacophore matching and structural similarity searches can then be used to investigate possible targets. Combining ligand-based and structure-based approaches can provide a broader view of potential drug-target interactions.
- Machine learning and AI have increasingly been incorporated into these workflows. Models can learn relationships between molecular structures, protein sequences, three-dimensional structures and experimental interaction data. AI methods can help prioritize potential targets, predict binding interactions or identify relationships within large drug-target datasets. However, predictions remain dependent on the quality, diversity and representativeness of the training data.
- Experimental validation remains essential. A computationally predicted drug-target interaction can be investigated using biochemical binding assays, enzymatic activity measurements, receptor assays, cellular experiments or biophysical techniques such as SPR, ITC, MST and BLI. Orthogonal experimental approaches are particularly valuable because different assays measure different aspects of molecular interaction and function.
- An important distinction should also be made between binding and functional activity. A drug may bind a protein without producing a significant functional change. Conversely, a ligand may produce a large functional effect through a relatively modest change in binding equilibrium if the target is part of an amplification pathway. Functional assays therefore provide information that cannot always be inferred from binding affinity alone.
- The cellular environment further complicates drug-target interactions. Proteins exist in crowded molecular environments and may form complexes with other proteins, undergo post-translational modifications or occupy specific cellular compartments. Protein abundance and localization can vary between tissues and disease states. Consequently, the interaction profile measured using purified proteins may differ from the effective target profile observed inside cells.
- This distinction is especially relevant for proteomics-based target identification. Techniques such as affinity-based proteomics, chemical proteomics and thermal-proteome approaches can investigate how compounds interact with proteins across complex biological samples. These methods can reveal unexpected targets that may not have been predicted from sequence or structural information alone.
- Drug repurposing provides another application of drug-target interaction networks. An approved or previously studied drug may interact with a protein or pathway that was not part of its original therapeutic mechanism. Computational target prediction, molecular docking, structural comparison, molecular dynamics and network analysis can generate hypotheses about such alternative mechanisms. Experimental studies are then required to determine whether the proposed interaction produces a meaningful biological effect.
- Polypharmacology is also important in understanding drug combinations. Two drugs may act on different targets within the same pathway, on parallel pathways or on completely different biological processes. Their combined effects can depend on the connectivity of the affected molecular networks. Computational drug-target networks can therefore provide a framework for investigating potential interactions between therapies, although biological validation remains necessary.
- The overall analysis of drug-target interactions can be organized into several interconnected layers. At the molecular level, researchers examine ligand structure and protein-ligand interactions. At the structural level, they investigate binding pockets, protein domains, conformational states and molecular dynamics. At the thermodynamic level, they estimate binding affinity using free-energy methods. At the cellular level, they examine target expression, localization and signaling pathways. At the systems level, they analyze networks of drug-target interactions and downstream biological responses.
- This creates a natural extension of the computational workflow developed throughout this series. Protein sequence analysis identifies candidate protein families and domains. Protein structure prediction, homology modeling or experimental structures provide three-dimensional models. Structural alignment and protein visualization reveal conserved and variable regions. Binding-site analysis identifies potential interaction sites. Molecular docking predicts possible ligand poses. Molecular dynamics simulations investigate conformational behavior. Binding free-energy calculations estimate relative or absolute energetic favorability. Finally, drug-target interaction analysis places these individual molecular events into the larger biological network.
- The same framework also highlights why no single computational method can completely describe drug action. Sequence similarity does not guarantee identical pharmacology. Structural similarity does not guarantee equivalent ligand affinity. Docking does not establish experimental binding. Binding affinity does not necessarily establish functional activity. And interaction with an individual protein does not necessarily explain the complete biological response of a drug.
- A comprehensive understanding of drug action therefore requires integration of molecular, structural, biochemical, cellular and systems-level information. This integrated perspective is particularly valuable in modern drug discovery, where researchers increasingly combine structural bioinformatics, molecular simulation, chemical informatics, proteomics, genomics and AI-based approaches.
- The progression can now be viewed as a hierarchy: protein sequence → protein family → protein domain → protein motif → domain architecture → protein structure → binding site → protein-ligand interaction → molecular docking → molecular dynamics → binding free energy → drug-target interaction → target selectivity → polypharmacology → cellular pathway response. Each level adds biological context to the molecular information obtained at the preceding level.
- Ultimately, drug-target interactions provide the connection between molecular recognition and pharmacological action. A drug’s behavior is determined not simply by whether it can bind a protein, but by the strength, duration and location of that interaction, the structural and dynamic properties of the target, the presence of related proteins, the cellular environment and the network of downstream pathways affected. Polypharmacology extends this concept further by recognizing that therapeutic and adverse effects can emerge from coordinated interactions with multiple molecular targets.
