Aqueous solubility and its structural determinants

A compound that will not dissolve cannot be dosed, cannot be assayed reliably, and frequently cannot be developed. Solubility is the property most often discovered too late, and it is the one where a structural fix has to be designed in rather than formulated around.

Thermodynamic solubility is the real one: the concentration of a saturated solution in equilibrium with the most stable solid form. Measuring it means shaking solid compound with buffer for a day or more and analysing the supernatant. It is slow and needs real material.

Kinetic solubility is what a screening cascade actually measures: dilute a DMSO stock into buffer and see what precipitates. It is fast, it needs micrograms — and it systematically overestimates, because the compound never had to leave a crystal lattice. It arrived already dissolved.

The gap between them can be a hundredfold. So when someone quotes a solubility, the first question is which measurement it was, and the second is whether the compound had ever been crystallised at that point.

Dissolving means doing two things, and they have different structural causes. The general solubility equation makes this explicit: solubility falls with lipophilicity and with melting point, separately.

• Escaping the crystal. The molecule has to leave the lattice. The cost is reflected in the melting point: a high-melting solid is held together well and dissolves badly.
• Being accommodated by water. Once free, the molecule has to be solvated. The cost here is reflected in logP: a greasy molecule is not welcome in water.

Recognising which of the two is your problem tells you what to do. A compound with logP 2 and a melting point of 250 °C has a crystal problem, and lowering logP further will not help. A compound with logP 6 and a melting point of 80 °C has a lipophilicity problem.

This is the single most useful structural insight in the module. Flat molecules stack face to face in the solid, symmetrical molecules pack efficiently in more orientations, and rigid molecules give up little entropy on crystallising. All three raise the melting point, and all three therefore hurt solubility.

So the classic insoluble compound is a flat, symmetrical, heavily aromatic molecule — which is unfortunately also the classic output of a project that has been optimising potency by adding aromatic rings.

Five moves, roughly in order of how reliably they work.

• Introduce an ionisable centre. The most effective by a wide margin. An ionised species is far more soluble, and it opens the door to salt forms. The cost is permeability, and possibly a new interaction you did not want.
• Break planarity. Twist two rings out of plane with an ortho substituent, or replace an aromatic ring with a saturated one. This attacks crystal packing directly.
• Disrupt symmetry. A methyl in the right place can lower a melting point by tens of degrees, by making efficient packing impossible. It is a cheap change with no cost in polarity.
• Raise the fraction of sp³ carbon. The module 12 point, applied here: three-dimensional molecules pack worse.
• Add a solubilising group. A morpholine, a piperazine, a short polyethylene glycol chain, hung off a position that points into solvent. Effective, but it adds weight and rotatable bonds, and it only works if there is a vector pointing out of the pocket.

Three things can be done after the molecule is fixed.

• Salt forms. If the compound is ionisable, make a salt. This is routine, well understood, and frequently sufficient. It requires an ionisable group, which is why the first structural lever above matters so much.
• Co-crystals. Crystallise the compound with a second, innocuous molecule that disrupts the packing. Works for neutral compounds, where a salt is not available.
• Prodrugs. Attach a solubilising group that is cleaved in vivo — a phosphate ester is the standard choice. Effective, and it adds a whole extra molecule to develop and register.

None of these fixes a molecule that is intrinsically too greasy. Formulation buys a factor of a few, occasionally ten. It does not buy three orders of magnitude, and a project relying on it to do so is usually a project that should have changed the molecule.

Finally, the reason this cannot be optimised in isolation.

To dissolve, a molecule wants to be polar and ionised. To cross a membrane passively, it wants to be neutral and lipophilic. These are opposite requirements, and a drug has to do both — dissolve in the gut, then cross the gut wall.

That is why the useful window is a window and not a direction. It is why logP 1 to 3 keeps appearing as a target rather than "as low as possible". And it is why optimising solubility alone produces compounds that dissolve beautifully and are never absorbed.