The hydrophobic effect and ordered water

Oil and water separate, and almost everyone explains it wrongly. There is no attraction between the oily bits. What drives them together is water getting its freedom back — and once you see binding that way, a binding site stops being an empty hole waiting for a ligand.

Put a non-polar solute in water. Water cannot hydrogen bond to it, so the water molecules around it reorganise: they form a more ordered cage-like shell, keeping their hydrogen bonds to each other by arranging themselves carefully around the intruder.

Those shell waters have lost entropy. They are more ordered than bulk water. That ordering is the real cost of dissolving something greasy.

Now bring two non-polar surfaces together. The shells between them are released back into bulk, and they regain their entropy. That gain is what drives association. The two surfaces are not attracted to each other in any special way — they are pushed together by water's preference for its own freedom.

So "hydrophobic bond" is a misnomer and worth correcting whenever you hear it. There is no bond. There is an entropically-driven exclusion.

The classical picture above is entropic, and at room temperature it largely is. But the split is temperature-dependent in a way that is a good test of whether you have understood the mechanism: as temperature rises the ordered shells are already disrupted by thermal motion, so there is less entropy to recover, and the driving force becomes progressively more enthalpic.

In real binding sites the picture is messier still. Many measured hydrophobic burials turn out to be enthalpically driven, because the water being released was not merely ordered but was making poor hydrogen bonds — which is the next section, and the practically useful one.

This is the sentence to take away from the module.

Before your ligand arrives, the pocket is full of water — usually an ordered network, often visible in the crystal structure as discrete well-behaved peaks in the density. Binding is not filling a hole. It is a competition: your ligand has to displace water that is already there, and it has to be worth more than the water was.

That reframing changes what you look for in a structure. The question is no longer "what contacts does my ligand make?" but "what was here before, and was it happy?"

Not all bound waters are equal, and the difference is exploitable.

A water in a polar, well-formed site makes its full complement of hydrogen bonds. It is comfortable, it is tightly held, and displacing it costs you everything it was gaining. Displacing a happy water is usually a bad trade.

A water in a hydrophobic or awkwardly-shaped pocket cannot make its hydrogen bonds. It is enthalpically frustrated — a high-energy or "unhappy" water. Displacing it releases it to bulk where it can bond properly, and you get that energy. This is close to free potency, and it is one of the more reliable structure-based design ideas there is.

The corollary is the design instruction: look for the frustrated waters and grow your ligand into them. A methyl or a small hydrophobic substituent aimed at an unhappy water is where several of medicinal chemistry's unexpectedly large potency jumps have come from — and it is another mechanism behind module 15's magic methyl.

The opposite case. Some waters are part of the binding site's architecture. They bridge ligand to protein, they appear in every structure of that target, and they are effectively a component of the surface your ligand recognises.

The classic is HIV-1 protease. A single water sits between the ligand and the two flap isoleucines, accepting from the flaps and donating to the ligand's carbonyls. Nearly every early protease inhibitor binds through it.

Most docking is done on a dry structure: waters stripped out, pocket treated as empty. That is a large approximation, and it fails in both directions.

• Strip a conserved structural water and you have deleted part of the binding site. Ligands that bind through it cannot be posed correctly.
• Keep every water and you block the pocket, so nothing docks.
• Keep the wrong subset and you bias the result in a way that is invisible afterwards.

There is no default that is right. The honest workflow is to look at the structure, decide deliberately which waters to keep, and record the decision so that the result can be interpreted later.

Methods do exist for doing better. Hydration site analysis — running a short simulation of the empty site and computing, for each water position, how much enthalpy and entropy it has relative to bulk — will tell you which waters are frustrated and worth displacing. It is more work than a docking run and it is the right tool when a project turns on this question.