A binding curve is a graph that shows how much of a molecule sticks to its target as you increase the molecule’s concentration. In its simplest form, the curve rises steeply at first, then levels off into a plateau as the target becomes saturated. That shape, and the precise point where the curve bends, encodes critical information about how tightly a drug grips its receptor, how efficiently an enzyme captures its substrate, or how readily hemoglobin picks up oxygen. The concept shows up everywhere from introductory biochemistry to billion-dollar drug-development programs, and the details of how a binding curve bends, steepens, or shifts can reveal far more than just “strong” or “weak” binding.
What a Binding Curve Actually Shows
Imagine you have a fixed number of target molecules, say a protein with a single binding pocket, and you gradually add more and more of a smaller molecule (a ligand) that fits into that pocket. At low concentrations, there is plenty of empty protein around, so nearly every ligand molecule finds a partner. The curve climbs sharply. As you keep adding ligand, free protein becomes scarce and the curve flattens: adding twice as much ligand no longer doubles the amount bound. Eventually the curve plateaus because every protein molecule is occupied.
The concentration of ligand at which exactly half the protein is occupied defines a key number called the dissociation constant, often written as Kd. A smaller Kd means tighter binding because you need less ligand to fill half the sites. At the halfway point on a standard saturation curve, the free ligand concentration equals the Kd.1Journal of Chemical Information and Modeling. Binding Curve Viewer: Visualizing the Equilibrium and Kinetics of Protein–Ligand Binding and Competitive Binding That single number anchors the entire curve: if you know the Kd and the total amount of protein present, you can predict how much complex will form at any given ligand concentration.
This classic shape, a smooth rise to a flat top, is called a rectangular hyperbola. It describes “simple” one-to-one binding where each protein has a single pocket and the ligand either fits or it doesn’t, with no influence from neighboring molecules. Many real-world interactions follow this pattern well enough for practical purposes, but some of the most biologically interesting systems deviate from it in revealing ways.
Why Some Curves Are S-Shaped
Hemoglobin is probably the most famous exception to the simple hyperbolic curve. When hemoglobin picks up its first oxygen molecule, its shape shifts in a way that makes the remaining binding sites grab oxygen more eagerly. The result is a sigmoidal, or S-shaped, binding curve: sluggish uptake at low oxygen levels, then a rapid surge, then the usual plateau. This behavior is called cooperative binding, and it occurs whenever the occupation of one site on a molecule changes the affinity of other sites on the same molecule. Cooperativity can be positive, meaning each binding event makes the next one easier, or negative, meaning each event makes the next harder.2PubMed Central. Cooperative binding
The steepness of the S is captured by a value called the Hill coefficient. A Hill coefficient of 1 means no cooperativity at all, producing the standard hyperbolic shape. A coefficient above 1 indicates positive cooperativity, and the higher the number, the more switch-like the response: the system goes from mostly unbound to mostly bound over a narrow range of ligand concentration. Hemoglobin’s Hill coefficient hovers around 2.8, which is why it loads up efficiently in the oxygen-rich lungs and unloads efficiently in the oxygen-poor tissues. That steep transition is the whole point of cooperativity in biology: it creates an on-off switch out of a graded chemical signal.
For comparison, myoglobin, the oxygen-carrying protein inside muscle cells, has only a single binding site and follows a simple hyperbolic curve. Myoglobin’s half-saturation point is much lower than hemoglobin’s, meaning it holds onto oxygen more tightly at low pressures, which is exactly what muscle tissue needs during exertion.3Applied Spectroscopy. Simultaneous Determination of Hemoglobin and Myoglobin Oxygen Binding Curves by Spectral Curve Fitting The contrast between hemoglobin’s sigmoidal curve and myoglobin’s hyperbola is a textbook illustration of how the shape of a binding curve, not just its position, determines biological function.
Binding Curves in Drug Development
When pharmaceutical researchers test how a drug affects cells, they typically expose cells to increasing drug concentrations and measure the response, usually something like cell survival or enzyme inhibition. The resulting plot is called a dose-response curve, and it is really a binding curve wearing different clothes. The concentration at which the drug achieves half its maximum effect, the IC50 or EC50, is the pharmacological cousin of the Kd.
For years, drug-screening programs focused almost exclusively on that halfway-point number, treating it as the single most important measure of a drug’s strength. But a large-scale analysis of cancer drug responses showed that other features of the curve matter just as much. Parameters like the steepness of the curve, the area under it, and the maximum effect the drug can achieve all varied systematically, and some of these features tracked with the drug’s class while others tracked with the cell line being tested.4Nature Chemical Biology. Metrics other than potency reveal systematic variation in responses to cancer drugs Two drugs might share the same IC50 but differ dramatically in how completely they shut down a target or how sharply their effect kicks in. A researcher looking only at potency would call them equivalent when they really aren’t.
This matters practically because the steepness of a dose-response curve tells you how precisely a drug can be dosed. A steep curve means a small change in concentration produces a large change in effect, which can be useful for an on-off therapeutic response but dangerous if the therapeutic window is narrow. A shallow curve is more forgiving of dose variability but harder to tune for a specific level of activity. Empirically, agonists acting at receptors are characterized by the maximum response they can produce, the concentration at which they reach half that maximum, and the Hill coefficient describing the curve’s slope.5Molecular Pharmacology. Dose–Response Analysis When There Is a Correlation between Affinity and Efficacy All three features are extracted from the same binding or dose-response curve, and all three influence clinical decisions.
How Inhibitors Reshape the Curve
When a second molecule competes for the same binding site as the original ligand, the binding curve shifts. In a competition assay, you measure how much of a labeled reference ligand gets displaced as you add increasing amounts of an unlabeled competitor. The resulting curve is an inverted sigmoidal: the signal starts high (lots of reference ligand bound), drops through a steep transition, and bottoms out when the competitor has kicked out nearly all the reference ligand.
What makes competition assays powerful is that different mechanisms of inhibition distort the curve in distinguishable ways. A competitive inhibitor, one that physically blocks the binding site, shifts the curve to the right: you need more reference ligand to achieve the same level of binding, but the maximum binding level stays unchanged. An uncompetitive inhibitor, which only binds after the ligand has already attached, changes the curve’s maximum rather than its midpoint. Mathematical models exist that can determine the mechanism and absolute potency of an inhibitor from a single dose-response experiment, provided the data are high quality.6PubMed Central. Equilibrium competition binding assay: inhibition mechanism from a single dose response Reading these curve shifts correctly is how researchers figure out not just how strong an inhibitor is, but how it works.
Measuring Binding Curves in the Lab
There are several mainstream techniques for generating binding data, and each has quirks that affect the shape and reliability of the resulting curves.
Isothermal titration calorimetry, or ITC, works by detecting the tiny amounts of heat released or absorbed when molecules bind. You inject small doses of ligand into a cell containing the target protein and measure the heat change after each injection. The resulting data, plotted as heat per injection versus the molar ratio of ligand to protein, form a binding isotherm. Fitting that isotherm yields the binding constant, along with the enthalpy and entropy contributions to binding, giving a thermodynamic portrait that no other single technique provides.7PubMed. Application of isothermal titration calorimetry in bioinorganic chemistry ITC’s main limitation is that it consumes relatively large amounts of material and can struggle with very tight binders whose binding events release too little measurable heat per injection.
Surface plasmon resonance, or SPR, takes a different approach. One binding partner is tethered to a sensor chip, and the other flows over it in solution. Binding changes the refractive index at the chip surface, producing a real-time trace of association and dissociation. A major advantage of SPR is that it separates the on-rate and off-rate of binding, which together determine the equilibrium constant but individually reveal how fast molecules find each other and how long they stay attached.8PubMed. Determination of rate and equilibrium binding constants for macromolecular interactions using surface plasmon resonance: use of nonlinear least squares analysis methods That kinetic dimension goes beyond what a static equilibrium curve can show, because a drug that binds slowly but dissociates even more slowly could be more effective in the body than a drug with the same Kd but fast on/off kinetics.9PubMed. Determination of kinetic data using surface plasmon resonance biosensors
Microscale thermophoresis, or MST, is a newer solution-based method. It tracks how fluorescently labeled molecules move through a temperature gradient, which changes when a binding partner is attached. Because both partners remain free in solution rather than pinned to a surface, MST avoids some artifacts that surface-based methods can introduce. Researchers have used it to determine binding affinities for protein-peptide interactions with good agreement to other methods.10PubMed. Microscale thermophoresis and fluorescence polarization assays of calcineurin-peptide interactions One practical challenge is that the fluorescent label itself can sometimes alter the binding interaction, and certain peptides can aggregate in ways that confuse the measurement.
When Targets Have More Than One Binding Site
The simple one-site model breaks down when a target molecule has two or more binding pockets. If those sites are identical and independent, the curve still looks hyperbolic, just reflecting the aggregate behavior. But if the sites differ in affinity or if binding at one site influences the other, the curve acquires features that a one-site model cannot explain: a subtle shoulder, a biphasic shape, or a Hill coefficient that deviates from 1.
Researchers handle this by fitting the data to multi-site models. For a protein with two distinct binding sites, models incorporate separate dissociation constants for each site and, if needed, a cooperativity factor describing how one site influences the other.11PubMed Central. Using two-site binding models to analyze microscale thermophoresis data Choosing the right model matters because forcing a two-site system into a one-site equation produces a Kd that is a meaningless average of the two real values. ITC data are particularly useful here because the heat signatures from the two sites often differ, giving the fitting algorithm something to grab onto.12PubMed Central. Fitting two- and three-site binding models to isothermal titration calorimetric data
This kind of complexity is not academic bookkeeping. Many therapeutically important targets, including ion channels, G-protein-coupled receptors, and multi-subunit enzymes, have multiple interacting binding sites. Getting the model wrong means getting the drug optimization wrong.
Avidity and the Antibody Problem
Antibodies present a special complication for binding curves because each Y-shaped antibody molecule has two identical binding arms. When both arms can latch onto neighboring copies of the same target at once, the effective binding strength, called avidity, can far exceed the intrinsic affinity of a single arm. The mechanism is straightforward: once one arm binds, the second arm is physically held close to its target, giving it a much higher chance of engaging. This forced proximity both delays the system from reaching true equilibrium and dramatically increases the time the antibody stays attached.13PubMed Central. Exploring avidity: understanding the potential gains in functional affinity and target residence time of bivalent and heterobivalent ligands
On a binding curve, avidity shows up as a steeper-than-expected rise, because the two-armed binding acts almost like a cooperative system even though no allosteric conformational change is involved. In engineered antibodies, the relationship between avidity and affinity can be tuned by adjusting how tightly each arm binds individually and how conditions like acidity modulate the interaction. Work on the IgG-FcRn interaction, which governs how long antibodies survive in the bloodstream, showed that an engineered antibody variant displayed an affinity inflection point around a specific pH and that avidity enhancement switched off at neutral pH, a designed feature that lets the antibody bind tightly inside acidic cellular compartments and release at normal blood pH.14PubMed Central. Insight into the avidity-affinity relationship of the bivalent, pH-dependent interaction between IgG and FcRn Understanding these curve features is essential for designing antibody therapies with the right duration of action.
Cooperativity Beyond Oxygen Transport
While hemoglobin is the most cited example, cooperative binding curves appear throughout biology in contexts that have nothing to do with blood. In gene regulation, transcription factors often bind DNA cooperatively, creating switch-like responses to changes in the concentration of a signaling molecule. When repressor proteins bind cooperatively, the transition from gene “on” to gene “off” becomes much sharper, with effective Hill coefficients around 3 compared to roughly 1.3 without cooperativity.15PubMed Central. A robust ultrasensitive transcriptional switch in noisy cellular environments That steepness is biologically useful because it filters out noise: small random fluctuations in repressor concentration don’t accidentally flip the gene on or off, but a genuine signal pushes through the threshold reliably.
The theoretical framework for understanding these cooperative systems dates back to the 1960s, with two influential models. One proposed that binding partners exist in a pre-existing equilibrium between active and inactive shapes, and ligand binding shifts that equilibrium. The other proposed that each binding event physically reshapes the molecule in a sequential fashion. Both models produce sigmoidal binding curves but make different predictions about intermediate states. Distinguishing between them experimentally remained difficult for decades, but modern computational and structural methods have renewed interest in understanding what cooperativity looks like at the atomic level.16PubMed Central. Allostery and cooperativity revisited
Equilibrium Versus Kinetics and Why It Matters for Drugs
A standard binding curve assumes the system has reached equilibrium, that molecules have had enough time to bind and unbind until a steady state is established. In a test tube with ample incubation time, this assumption holds well. In a living body, it often doesn’t. Drug concentrations in tissues fluctuate as the drug is absorbed, distributed, and eliminated. If a drug binds slowly, it may never reach equilibrium concentrations at its target before being cleared. If it dissociates slowly, it can continue working long after its concentration in the blood has dropped.
This realization has pushed the drug industry to pay attention to binding kinetics, not just equilibrium constants. A considerable number of approved drugs show non-equilibrium binding characteristics, suggesting that how long a drug stays attached to its target, its residence time, matters for how well it works in practice.17PubMed. Kinetics for Drug Discovery: an industry-driven effort to target drug residence time Two drugs with identical Kd values can have vastly different clinical profiles if one binds and releases quickly while the other binds and stays put. The binding curve alone, which captures equilibrium behavior, cannot distinguish between these two scenarios. That is why SPR-based kinetic measurements, which separately resolve on-rates and off-rates, have become standard in drug-discovery pipelines.
Separating Specific From Non-Specific Binding
Every experimental binding curve contains some noise from non-specific binding, which is the tendency of molecules to stick weakly and nonselectively to surfaces, container walls, or irrelevant sites on the target. At low concentrations of ligand, specific binding dominates and the curve rises steeply. At high concentrations, specific sites are saturated and non-specific binding, which increases linearly with concentration rather than plateauing, becomes a larger fraction of the total signal. If you don’t account for it, the curve looks like it never fully levels off, and the Kd you extract will be wrong.
The standard approach is to run a parallel experiment with a known competitor that blocks only the specific binding site. Subtracting the competitor-blocked signal from the total signal gives the specific binding curve. In one study using microscale thermophoresis to measure a receptor-ligand interaction, the Kd from the total-binding curve was about 4.3 nM, while the Kd after subtracting non-specific binding was about 5.3 nM, a reassuringly small difference in that case but one that can be much larger for weaker or messier interactions.18Scientific Reports. An original approach to measure ligand/receptor binding affinity in non-purified samples Neglecting this correction is one of the more common sources of unreliable binding data in the literature.
Enthalpy-Entropy Compensation and the Thermodynamic Binding Curve
When researchers use ITC to dissect binding thermodynamically, they often encounter a frustrating pattern. Making a chemical modification to a drug that strengthens one energetic component of binding, say the enthalpic contribution from hydrogen bonds, frequently weakens another component, the entropic contribution from water displacement or molecular flexibility. This tug-of-war, called enthalpy-entropy compensation, means that large improvements in affinity are harder to achieve than you’d expect from optimizing one parameter at a time.
An analysis across 32 diverse proteins found that strong compensation, where the enthalpic and entropic contributions nearly cancel each other out, occurred in roughly a fifth of chemical modifications. On the other hand, about 15% of modifications showed reinforcement, where both components moved in the same favorable direction.19PubMed Central. Extent of enthalpy-entropy compensation in protein-ligand interactions The practical lesson for drug designers is that a binding curve showing improved affinity after a chemical tweak may be hiding compensatory thermodynamic shifts underneath. Two drugs with identical Kd values can have completely different thermodynamic profiles, and those profiles can influence properties like selectivity and off-target binding that the Kd alone doesn’t capture.
Tight-Binding Complications
When a ligand binds its target extremely tightly, with a Kd in the low-nanomolar or picomolar range, the standard assumptions behind binding-curve analysis start to break down. The core assumption is that the free ligand concentration approximately equals the total ligand concentration. For tight binders at the concentrations typically used in assays, a substantial fraction of the ligand gets sequestered in the bound complex, so the free concentration is much lower than what was added to the tube. The binding curve in this regime looks almost like a straight line rising to a sharp corner and then flattening, rather than a smooth hyperbola. Extracting a reliable Kd from that shape requires specialized fitting approaches and careful attention to assay design, particularly the concentrations of target and ligand relative to the expected Kd.20PubMed Central. Quantifying the Interactions between Biomolecules: Guidelines for Assay Design and Data Analysis
Getting this wrong is surprisingly easy. If your protein concentration is 100-fold above the Kd, the binding curve becomes a titration endpoint: you see a straight line to a stoichiometric breakpoint, and the curve contains almost no information about binding strength. The fix is to lower the protein concentration, but that requires a detection method sensitive enough to measure binding at those lower concentrations. This tension between sensitivity and the physics of tight binding is one of the ongoing practical challenges in biophysics labs.
Computational Prediction of Binding Curves
With the rise of machine learning, there has been a surge of effort to predict binding affinities, and by extension binding curves, computationally rather than experimentally. Deep learning models trained on large databases of protein-ligand structures attempt to predict Kd or IC50 values directly from molecular structure.21Briefings in Bioinformatics. Prediction of protein–ligand binding affinity via deep learning models If successful, these tools could dramatically accelerate drug discovery by narrowing down candidate molecules before any wet-lab experiments.
The reality, though, is more humbling. Predicting absolute binding affinities from structure alone remains extremely difficult because the free energy differences that determine tight versus weak binding are small, on the order of a few kilocalories per mole, while the total energies involved are enormous. Current computational methods are generally better at ranking compounds, telling you which of several candidates is likely to bind more tightly, than at predicting absolute Kd values. They also struggle with the thermodynamic compensation effects described earlier, where improving one part of the interaction worsens another. For the foreseeable future, experimental binding curves remain the gold standard, with computation serving as a useful filter rather than a replacement.