Plasma protein binding and the free drug hypothesis

A docking score tells you how tightly a molecule binds its target in an idealized pocket. What it cannot tell you is how much of that molecule will ever be free in the bloodstream to reach the target at all. Most of a circulating drug is stuck to plasma proteins, and only the unbound fraction does the work. Understanding that distinction, and the ways it is routinely misused, separates a useful potency number from a misleading one.

What plasma protein binding actually is

Once a drug enters the blood, it partitions between two pools: bound to plasma proteins, and free in solution. The dominant binders are albumin (which carries acidic and neutral drugs) and alpha-1-acid glycoprotein (which carries many basic drugs). Binding is reversible and fast, so the two pools stay in equilibrium. The parameter that captures it is the fraction unbound, written fu: the ratio of free drug concentration to total drug concentration. A drug that is 99% bound has an fu of 0.01.

The free drug hypothesis

The governing principle is simple: only unbound drug crosses membranes, distributes into tissue, and engages the target. Protein complexes are too large and too transient to act on the receptor. At steady state, the free concentration in plasma equilibrates with the free concentration at the site of action, so the unbound plasma level is the quantity that matters for efficacy.

The practical consequence is that potency and exposure must be compared on the same footing. An in vitro IC50 is measured in a protein-poor buffer, so it is effectively a free-drug potency. To ask whether a dose will work, you compare that free IC50 against the free (not total) plasma concentration the dose produces. Comparing a free IC50 against a total plasma level, ignoring the 99% that is bound, is one of the most common ways a program fools itself into thinking a compound is more potent in vivo than it really is.

The trap: optimizing protein binding for its own sake

It is tempting to read “99% bound” as a problem to be engineered away, on the theory that lowering protein binding frees up more drug. This is usually a mistake, and it is worth being explicit about why.

For an orally dosed drug at steady state, the average free concentration is set by the dose rate and the unbound intrinsic clearance, not by the fraction bound. If you reduce protein binding, you free up more drug, but you also expose more drug to clearance, and the two effects cancel. The free concentration, the thing that drives efficacy, stays put. So chemically chasing a lower protein binding number generally moves a lot of structure for no pharmacological gain, and risks degrading the properties that actually mattered.

The better framing: protein binding is a scaling factor you measure so you can interpret total concentrations correctly, not a property you optimize. The levers worth pulling are unbound potency and metabolic stability.

Where fu genuinely matters

• Interpreting PK/PD. You cannot build a sound exposure-response relationship without converting total drug levels to free levels using fu.
• Drug-drug interactions. Displacement from albumin is often overstated, but fu still feeds the calculations that estimate interaction risk and therapeutic index.
• Very highly bound compounds. When fu drops below a few percent, fu is hard to measure accurately, and small measurement errors translate into large errors in the predicted free concentration. For these molecules, predicting an efficacious human dose from in vitro potency and fu alone is unreliable and should be treated with caution.

How this connects to docking

Docking and the free drug hypothesis live at opposite ends of the same pipeline. Docking estimates binding affinity to the target, which corresponds to the unbound potency you would measure in a clean assay. Plasma protein binding then governs how much of your compound is available to realize that affinity in a living system. A molecule can dock beautifully and still fail in vivo if it is so highly and non-specifically bound that the free concentration never reaches its own IC50. Keeping the two ideas distinct (target affinity from docking, availability from fu) is what stops a strong docking result from being over-read.

Try the docking yourself

Open Studio and dock a candidate against your target to get the affinity side of the picture, then read the predicted score as an unbound potency, the number you would later compare against a free plasma concentration rather than a total one. Liganx puts molecular docking online and free in the browser, so you can run molecular docking on a series and rank by target affinity before the ADMET properties like protein binding decide which of them survives in vivo.

Primary sources

• Smith DA, Di L, Kerns EH. The effect of plasma protein binding on in vivo efficacy: misconceptions in drug discovery. Nat Rev Drug Discov 9, 929–939 (2010). doi:10.1038/nrd3287
• Liu X, et al. Unbound drug concentration in brain homogenate and cerebral spinal fluid at steady state. J Pharm Sci 106, 2475–2485 (2017). doi:10.1016/j.xphs.2017.04.018
• Di L. Free drug concepts: a lingering problem in drug discovery. J Med Chem (2025). doi:10.1021/acs.jmedchem.5c00725