Molecular recognition · an interactive field guide

Why does one molecule stick while another slips away?

Follow a small drug molecule as it finds a protein pocket—from fluctuating electron clouds to whole-population equilibrium. By the end, you will be able to separate getting there, fitting, making contacts, and staying bound.

The kernel

At the bottom, binding is one quantum system finding a lower free-energy arrangement—but that single result is best understood by peeling it into electron interactions, water rearrangement, and molecular motion.

01 · First principles

When atoms approach, they become one quantum problem.

Each nucleus attracts every electron; electrons repel one another; nuclei repel one another; and confining an electron’s wavefunction costs kinetic energy. The Schrödinger equation balances all four at once. Far apart, “an orbital on atom A” is a useful approximation. Close together, the stationary states extend across both partners and the electron density rearranges.

The animation uses the simplest honest picture: one electron shared by two nuclei. Moving the nuclei changes the combined wavefunction. More density can accumulate between them, but nuclear repulsion eventually prevents collapse.

Two nuclei sharing a one-electron stateanimate · approach and separate
From Schrödinger’s equation to DFT
For fixed nuclei, \(\hat H\Psi=E\Psi\). The electronic Hamiltonian contains electron kinetic energy, electron–nucleus attraction, electron–electron repulsion, and nucleus–nucleus repulsion: \[ \hat H=-\sum_i\frac{\hbar^2}{2m_e}\nabla_i^2-\sum_{iA}\frac{Z_Ae^2}{4\pi\epsilon_0r_{iA}}+\sum_{i<j}\frac{e^2}{4\pi\epsilon_0r_{ij}}+\sum_{A<B}\frac{Z_AZ_Be^2}{4\pi\epsilon_0R_{AB}}. \] \(\Psi\) is a joint amplitude over all electron coordinates, not one independent wave per atom. The Born–Oppenheimer approximation solves this electronic problem at each nuclear geometry to obtain a potential-energy surface. Wavefunction methods approximate \(\Psi\); density-functional theory (DFT) instead seeks the ground-state energy as a functional of the electron density \(n(\mathbf r)\). Neither makes binding a separate force: binding is an energy difference between geometries.
02 · Van der Waals contact

Neutral clouds attract at long range and resist overlap at short range.

Electron density fluctuates even in a neutral atom. One fleeting dipole polarizes its neighbor; quantum mechanics correlates the fluctuations so attractive arrangements occur slightly more often. This is London dispersion. At short range, occupied states must remain antisymmetric under electron exchange. Forcing overlapping electrons into distinguishable states adds nodes and kinetic energy: exchange, or Pauli, repulsion.

Correlated clouds meeting the Pauli wallanimate · follow both terms
Where the familiar van der Waals formula comes from
Second-order perturbation theory for two polarizable quantum systems gives the leading dispersion term \(U_\mathrm{disp}(r)=-C_6/r^6\). \(C_6\) depends on their excitation spectra and polarizabilities. Exchange repulsion decays roughly exponentially, \(U_\mathrm{ex}\approx Ae^{-br}\). Force fields often replace that combination with the convenient Lennard–Jones form \(4\varepsilon[(\sigma/r)^{12}-(\sigma/r)^6]\); its \(r^{-12}\) wall is computational convenience, not a first-principles law.
03 · Coulomb interactions and polarization

“Neutral” molecules contain electrical landscapes.

Nuclei and electron density rarely cancel point-by-point. A polar bond leaves partial positive and negative regions; several bonds create dipoles and higher multipoles. Their electric fields act through ordinary Coulomb attraction and repulsion. The field also distorts the neighbor’s electron density—polarization—which is usually attractive because the induced density relaxes into the field.

A molecular dipole rotating in a protein electric fieldanimate · rotate the dipole
Charges, multipoles, dielectric screening, and induction
For point charges, \(U=\sum_{ij}q_iq_j/(4\pi\epsilon_0\epsilon_r r_{ij})\). Partial charges approximate the electrostatic potential produced by a continuous quantum density. At longer range, a dipole \(\boldsymbol\mu\) in a field \(\mathbf E\) has \(U=-\boldsymbol\mu\cdot\mathbf E=-\mu E\cos\theta\). A polarizable site develops \(\boldsymbol\mu_\mathrm{ind}=\alpha\mathbf E\), giving \(U_\mathrm{ind}=-\tfrac12\alpha E^2\). The factor \(1/2\) appears because the dipole grows from zero while the field does work. Water and ions screen fields, so the local protein dielectric environment matters.
04 · Hydrogen-bond geometry

A hydrogen bond is an electrostatic contact sharpened by quantum directionality.

A donor bond \(D{-}H\) exposes a partially positive hydrogen. An acceptor carries negative potential and usually a directed lone-pair orbital. Electrostatics brings them together; polarization and slight donation from the occupied lone pair into the donor’s empty antibonding orbital reinforce geometries where \(D{-}H\cdots A\) is close to straight. Bend or stretch the contact and those contributions weaken.

A lone pair aligning with a donor antibondanimate · bend the contact
What “orbital interaction” means here
A common donor–acceptor analysis describes mixing of the acceptor lone pair \(n_A\) with the donor bond’s antibonding orbital \(\sigma^*_{D-H}\). In second-order perturbation language, stabilization grows roughly like \(-|\langle n_A|\hat H|\sigma^*\rangle|^2/\Delta E\): larger spatial overlap and a smaller orbital-energy gap strengthen it. The widget’s \(\cos^2\theta\) factor is a geometric toy, not a universal hydrogen-bond law. Energy decomposition methods disagree on exact percentages because electrostatics, polarization, and charge transfer are not uniquely separable.
05 · Solvent exposure and desolvation

Binding partners arrive already bound to water.

A polar group in solution is stabilized by oriented water molecules. To make a direct protein–drug contact, those waters must leave; the new contact must repay that desolvation cost. Burying a polar group without a partner is therefore expensive. Nonpolar surfaces pose a different problem: nearby water has fewer comfortable arrangements. Burying those surfaces together can release constrained water back to the bulk—the hydrophobic effect.

Water shells leaving as two surfaces meetanimate · bind through solvent
A thermodynamic cycle for solvent
Imagine first removing each partner from water, then letting the dry partners interact, then allowing water to relax around the complex. The binding free energy is unchanged because free energy is a state function. This cycle exposes the competition: \[ \Delta G_\mathrm{bind}=\Delta G_\mathrm{desolv,P}+\Delta G_\mathrm{desolv,L}+\Delta G_\mathrm{direct}+\Delta G_\mathrm{solvent\ release}. \] Continuum-solvent models estimate polar electrostatics and nonpolar surface terms; explicit-water simulations represent individual water molecules. Neither “hydrogen bond good” nor “hydrophobic contact good” is meaningful without asking what interactions and water configurations were lost.
06 · Conformational preferences

The binding pose must be paid for before it can be rewarded.

A flexible drug and protein each occupy an ensemble of shapes. If only a rare shape fits, binding must first select that shape; its rarity is a free-energy cost. Bending bonds or torsions away from their preferred values adds strain. Once bound, translations, rotations, and internal motions are also restricted. Preorganization helps because the free molecule already spends more time near the binding-ready shape.

An ensemble searching for its binding-ready shapeanimate · cycle conformers
From populations to conformational free energy
If the binding-compatible region has equilibrium probability \(p_\mathrm{ready}\) in the free ensemble, selecting it costs \(\Delta G_\mathrm{select}=-RT\ln p_\mathrm{ready}\). For \(N\) equally probable conformers with only one ready, this becomes \(RT\ln N\). Real conformers are not equally probable: their weights follow \(p_i=e^{-G_i/RT}/Z\), where \(Z=\sum_i e^{-G_i/RT}\). “Conformational selection” and “induced fit” describe kinetic routes; equilibrium affinity depends on the free energies of the full free and bound ensembles.
07 · The complete ledger

Binding is the sum over everything that changes.

Now bring two ligand atoms toward a three-atom pocket. No single curve decides the result. Dispersion, electrostatics, polarization, and a correctly aligned hydrogen bond lower the interaction energy. Pauli overlap, desolvation, and conformational restriction oppose them. The total develops a well only when one geometry makes the entire ledger favorable.

Five atoms building a binding free-energy wellanimate · sum every layer
Pauli + strainelectrostatics + H bonddispersion + polarizationsolventtotal
From microscopic energy to measurable affinity
A force-field score at one pose is not yet a binding free energy. Statistical mechanics integrates over all solvent, protein, and ligand configurations: \[ \Delta G^\circ_\mathrm{bind}=-RT\ln\!\left(\frac{Z_{PL}}{Z_PZ_L}C^\circ\right). \] The partition functions \(Z\) sum Boltzmann weights over quantum states and nuclear configurations. In practice, molecular dynamics uses an approximate potential—often fitted to quantum calculations—and free-energy methods sample the configurations. The result connects to experiment through \(\Delta G^\circ=RT\ln(K_D/1\,\mathrm M)\). A favorable pose is necessary; a favorable ensemble average is what produces affinity.
Putting it together

There is one physics, viewed at useful resolutions.

At the finest layer, nuclei and electrons form one quantum state. That state’s energy changes through electrostatics, correlation, polarization, orbital mixing, and exchange. At the molecular layer, those effects appear as van der Waals contacts, hydrogen bonds, charge complementarity, and strain. In solution, water and fluctuating conformations add enormous numbers of competing arrangements. Binding occurs when the whole bound ensemble has lower free energy—not when a shape-matching rule fires.

DFT can estimate the electronic energy of a small frozen geometry. Molecular mechanics makes larger systems affordable. Molecular dynamics samples their motion. Free-energy perturbation, thermodynamic integration, and related methods turn those samples into \(\Delta G\). Experiment measures the final population ratio as \(K_D\), \(K_i\), or a kinetic pair \(k_\mathrm{on}\) and \(k_\mathrm{off}\).