Noncovalent Interactions in Aqueous Systems

A protein folds into a working shape in milliseconds. It is held together by forces individually weaker than a single covalent bond. There are no disulfide crosslinks, no permanent chemistry. Instead there are thousands of transient contacts, each worth a few kilojoules per mole. These sum to something stable enough to catalyze reactions at diffusion-limited rates. Yet the result stays flexible enough to unfold, refold, and turn over. Understanding how life builds robust structure from weak, reversible forces is one of the deepest questions in biochemistry. The answer lies in comparing four qualitatively distinct interactions — ionic, hydrogen-bonding, van der Waals, and hydrophobic. It also lies in seeing how water itself decides which of them matter. These four forces are not competitors of equal standing; water rigs the game. An ionic bond worth 20 kJ/mol in vacuum collapses to 2–3 kJ/mol once both charges are hydrated. This happens because water's own dielectric screening does most of the work the "bond" was supposedly doing. A hydrogen bond gains almost nothing thermodynamically when it forms inside a folded protein. It merely replaces a hydrogen bond to water that existed in the unfolded state. Van der Waals contacts are individually feeble (0.1–1 kJ/mol). But they are so numerous in a tightly packed protein interior that their sum rivals covalent bond strengths. And the hydrophobic effect is the single largest term driving folding and binding in water. It is not really an "interaction" between nonpolar groups at all. Instead, it is an entropic tax that water charges for keeping oily surface exposed. That tax is refunded when the surface buries itself. The recurring theme is context-dependence. The same chemical group can contribute strongly, weakly, or oppositely to binding. This depends on whether its partner is water or another molecule. This forces a shift in thinking from bond energies, an intrinsic, additive quantity, to free energies of transfer, a comparative, solvent-referenced quantity. Biochemistry is fundamentally a science of differences. It is the difference between a group's environment before and after folding, binding, or catalysis. Nowhere is that clearer than in comparing the four noncovalent forces side by side. This article builds a quantitative, mechanistic comparison. It covers the physical origin of each force, its distance and orientation dependence, and its magnitude in vacuum versus in water. It includes worked calculations showing how much any single interaction actually contributes to a real biological free energy of binding. It closes by connecting weak, additive noncovalent chemistry to the entropic bookkeeping. That bookkeeping governs why life runs as an engine at all.

Key takeaways

Ionic interactions in water are governed by Coulomb's law screened by a dielectric constant of ~78, cutting vacuum bond strengths by roughly 40-fold.
Hydrogen bonds are largely enthalpically "invisible" in folding because they trade a bond to water for a bond to a partner of similar strength.
Van der Waals (London dispersion) forces scale as $r^{-6}$ and are individually tiny (~0.4–4 kJ/mol) but massively additive over a close-packed interface.
The hydrophobic effect is entropy-driven at room temperature — it comes from releasing ordered water, not from any attraction between nonpolar atoms.
All four forces are additive in free energy, but not in an intuitive "add the bond energies" sense — each must be referenced to the correct unbound/solvated state.
Real binding free energies (5–15 kcal/mol, i.e., ~20–60 kJ/mol) emerge from partial cancellation of large opposing terms, not from simple summation of maximum bond strengths.
Key idea: every noncovalent force must be measured as a difference between two solvated states, not as an intrinsic bond energy, because water competes for every one of them.

Where this lives today

Structure-based drug design leans on exactly this hierarchy of forces. Modern scoring functions in computational docking decompose predicted binding free energy into separately weighted electrostatic, hydrogen-bond, van der Waals, and solvation/hydrophobic terms, precisely mirroring the additive decomposition worked through above; medicinal chemists routinely trade a lost hydrogen bond for improved shape complementarity (more van der Waals contact) when optimizing a lead compound, because the magnitudes are genuinely comparable. Fragment-based drug discovery specifically hunts for small molecules that maximize van der Waals packing efficiency per atom, since gross hydrogen-bonding potential is easily squandered by desolvation penalties if geometry is imperfect. Protein engineering and de novo protein design (the Rosetta and, more recently, deep-learning-based design pipelines) explicitly optimize buried hydrophobic surface area and close-packed van der Waals cores as the primary stability determinants, adding hydrogen bonds and salt bridges mainly for specificity and solubility rather than for raw stability, a direct engineering consequence of the small net per-hydrogen-bond value derived here. Membrane biophysics depends on the same hierarchy in reverse: the low dielectric constant of the lipid bilayer interior ( $\varepsilon_r \approx 2$ –4) makes ionic interactions and unpaired charges catastrophically unfavorable inside membranes, which is precisely why integral membrane proteins bury charged residues only when paired in salt bridges or coordinated by multiple polar groups, and why the "positive-inside rule" and hydrophobic transmembrane helix recognition algorithms used in every membrane-protein topology predictor are built directly on the dielectric-screening physics developed above. Ion channel selectivity filters (e.g., the KcsA potassium channel) achieve their remarkable ion selectivity by using backbone carbonyl oxygens to mimic the hydration shell of K⁺ almost exactly in geometry and number, a direct engineering application of the competition between ion–water and ion–protein electrostatic interactions quantified in Derivation 1. DNA-binding protein design and CRISPR-associated protein engineering exploit the polyelectrolyte counterion-release effect quantified in Worked Example 1, where entropically favorable release of condensed Na⁺/Mg²⁺ counterions from DNA phosphates upon protein binding contributes as much binding free energy as direct electrostatic contacts. Cryo-EM and X-ray structure refinement software use exactly the Lennard-Jones and Coulomb-with-dielectric terms above as the physical basis of their energy functions when validating and refining atomic models against experimental density.

Ask the tutor: How does the balance of these four forces shift when a protein binds RNA versus DNA, or when it operates in the unusually crowded, low-water-activity environment of the cytoplasm?

The physical picture

Consider two charged groups, a hydrogen-bond donor/acceptor pair, and two greasy side chains, all initially fully solvated by water, then brought together to form a folded protein or a ligand-receptor complex. Four physically distinct forces operate simultaneously, but each has a completely different relationship to the solvent.

Ionic interactions are pure electrostatics: Coulomb's law between point charges, attenuated by the dielectric constant of the intervening medium. Water's enormous dielectric constant ( $\varepsilon_r \approx 78.5$ at 25 °C) arises because its permanent dipole moment (1.85 D) reorients in response to any field, screening charges far more effectively than a nonpolar solvent ( $\varepsilon_r \approx 2$ –4) or vacuum ( $\varepsilon_r = 1$ ) could. This screening is why a buried salt bridge inside a protein, shielded from bulk water, can be far stronger than the same pair of charges sitting on the surface.

Hydrogen bonds are directional, quantum-mechanically real but classically well-approximated as electrostatic-plus-partial-covalent interactions between a polarized X–H donor and a lone pair acceptor Y. Their strength (10–40 kJ/mol depending on donor/acceptor and geometry) depends steeply on the X–H···Y angle (near-linear is optimal) and on the H···Y distance. Crucially, liquid water itself is a dense three-dimensional hydrogen-bond network — every polar group not tied up in a biomolecular H-bond is instead hydrogen-bonded to solvent. This means the net energetic gain from forming an intramolecular H-bond in a folded protein is the difference between (H-bond to partner) and (H-bond to water lost), a much smaller number than the gross bond strength.

Van der Waals interactions comprise the induced dipole–induced dipole attraction (London dispersion), always present between any two atoms, plus a short-range Pauli/exchange repulsion once electron clouds overlap. The balance defines the Lennard-Jones potential, with a shallow minimum ( $\varepsilon \sim 0.1$ –1 kJ/mol per atom pair) at the van der Waals contact distance. Individually negligible, these forces become significant only through close, complementary shape-matching that lets thousands of atom pairs sit simultaneously near their energy minima — which is precisely what enzyme active sites and protein cores are built to do.

The hydrophobic effect stands apart: it is not a direct force between nonpolar atoms at all (their mutual van der Waals attraction is unremarkable) but a consequence of what water must do around them. Water molecules adjacent to a nonpolar surface cannot form hydrogen bonds "outward" into the solute, so they reorganize into a partially ordered cage (clathrate-like for small nonpolar solutes) that preserves hydrogen bonding among themselves at some entropic cost. Removing that surface from water — by burying it against another nonpolar surface — releases the constrained water back into the bulk, increasing entropy. At physiological temperature this entropy gain dominates the free energy of transfer of a hydrocarbon from water into oil.

Figure

Four noncovalent forces: distance dependence and typical strength in water

Figure

Salt bridge free energy: vacuum versus water, and buried versus exposed

The core ideas

Coulomb's law with dielectric screening, governing ionic interactions:

E_{\text{ionic}} = \frac{q_1 q_2}{4\pi\varepsilon_0 \varepsilon_r r}

where: $q_1, q_2$ are the point charges (C), $\varepsilon_0 = 8.854\times10^{-12}\ \text{F/m}$ is the vacuum permittivity, $\varepsilon_r$ is the relative dielectric constant of the medium (dimensionless), and $r$ is the charge separation (m). Reading: the force between two charges falls off as $1/r$ (energy) — long-ranged compared to the other three forces — but is divided by $\varepsilon_r$ , which is ~78.5 in bulk water and can be as low as 2–4 in a protein interior or membrane. This single parameter is why ionic interactions are so wildly context-dependent.

Hydrogen-bond geometry dependence, an empirical/electrostatic model:

E_{\text{HB}}(r,\theta) \approx E_{\text{HB}}^{0}\left(\frac{r_0}{r}\right)^{n}\cos^{2}\theta

where: $E_{\text{HB}}^0$ is the optimal bond energy at ideal distance $r_0$ (~2.8–3.0 Å heavy-atom separation) and linear geometry, $\theta$ is the deviation from linearity of the D–H···A angle, and $n$ is typically 2–4 depending on the model used for the distance falloff. Reading: hydrogen bonds are steep in both distance and angle — bending 30° off-axis can cut the bond energy by half — which is why hydrogen bonds impose strict directional constraints on secondary structure (α-helix, β-sheet) that pure electrostatics would not.

The Lennard-Jones potential, for van der Waals interactions:

E_{\text{vdW}}(r) = 4\varepsilon\left[\left(\frac{\sigma}{r}\right)^{12} - \left(\frac{\sigma}{r}\right)^{6}\right]

where: $\varepsilon$ is the depth of the potential well (energy scale, ~0.1–1 kJ/mol per atom pair), $\sigma$ is the finite distance at which the potential is zero, and $r$ is the interatomic distance. Reading: the $r^{-6}$ attractive term is the London dispersion force from correlated instantaneous dipoles; the $r^{-12}$ term is a steep empirical stand-in for Pauli exclusion repulsion. The minimum sits at $r_{\min}=2^{1/6}\sigma$ , defining the van der Waals contact distance.

The hydrophobic transfer free energy, empirically linear in buried nonpolar surface area:

\Delta G_{\text{transfer}} = \gamma \, \Delta A_{\text{nonpolar}}

where: $\gamma$ is an empirical surface-tension-like coefficient (roughly 25–100 J mol⁻¹ Å⁻² depending on the reference data set and temperature), and $\Delta A_{\text{nonpolar}}$ is the nonpolar surface area removed from water contact (Å²). Reading: this is not a pairwise atom–atom potential like the other three; it is a bulk solvent reorganization free energy, proportional to the area of oil–water interface eliminated, and decomposes into large, partially compensating enthalpy and entropy terms that individually vary strongly with temperature even though $\Delta G$ itself is fairly flat. The full free energy of association combining all four contributions is

\Delta G_{\text{binding}} = \Delta G_{\text{ionic}} + \Delta G_{\text{HB}} + \Delta G_{\text{vdW}} + \Delta G_{\text{hydrophobic}} - T\Delta S_{\text{conf}}

where the final conformational entropy term (loss of translational, rotational, and side-chain rotameric freedom on binding) is a large unfavorable term working against all four favorable contributions — set up here as a bridge to the more general entropic accounting of cellular free-energy transactions.

Derivation ADielectric screening of a buried salt bridge — from vacuum Coulomb energy to the true aqueous value

Goal. Show quantitatively why a salt bridge quoted in textbooks as "20 kJ/mol" is really only a few kJ/mol in water, and why burial changes this dramatically.

Step 1 — Set up the unscreened Coulomb energy. Take a Arg⁺ guanidinium nitrogen and an Asp⁻ carboxylate oxygen at a typical hydrogen-bonded salt-bridge distance $r = 0.30\ \text{nm} = 3.0\times10^{-10}\ \text{m}$ , each carrying charge magnitude $e = 1.602\times10^{-19}\ \text{C}$ (treating each as a unit point charge, a simplification but standard for order-of-magnitude work). In vacuum ( $\varepsilon_r=1$ ):

E_{\text{vac}} = \frac{e^2}{4\pi\varepsilon_0 r} = \frac{(1.602\times10^{-19})^2}{4\pi(8.854\times10^{-12})(3.0\times10^{-10})}

Evaluating the numerator: $(1.602\times10^{-19})^2 = 2.566\times10^{-38}\ \text{C}^2$ . Evaluating the denominator: $4\pi(8.854\times10^{-12})(3.0\times10^{-10}) = 3.341\times10^{-19}$ .

E_{\text{vac}} = \frac{2.566\times10^{-38}}{3.341\times10^{-19}} = 7.68\times10^{-19}\ \text{J per ion pair}

Converting to a molar quantity by multiplying by Avogadro's number $N_A = 6.022\times10^{23}\ \text{mol}^{-1}$ :

E_{\text{vac,molar}} = (7.68\times10^{-19})(6.022\times10^{23}) = 4.63\times10^{5}\ \text{J/mol} = 463\ \text{kJ/mol}

This is the number that would apply if the two ions sat in a vacuum — comparable to a covalent bond, and never actually realized biologically.

Step 2 — Screen by bulk water's dielectric constant. Water's relative permittivity at 25 °C is $\varepsilon_r = 78.5$ . Dividing:

E_{\text{water}} = \frac{463\ \text{kJ/mol}}{78.5} = 5.90\ \text{kJ/mol}

This collapse — from 463 to 5.9 kJ/mol, a factor of 78.5 — is the central lesson: bulk water annihilates roughly 98.7% of the naive Coulombic attraction. A surface salt bridge on a protein, fully solvent-exposed, contributes only this much (and even less once the entropic cost of ordering water around the exposed charges and the competition with ion pairing to free Cl⁻/Na⁻ in solution are included; measured surface salt-bridge stabilization is often only 1–3 kJ/mol).

Step 3 — Recompute for a buried salt bridge in a low-dielectric protein interior. Protein interiors, being densely packed with polarizable but non-reorienting groups (no bulk dipolar reorientation as in water), are typically modeled with $\varepsilon_r \approx 4$ :

E_{\text{buried}} = \frac{463\ \text{kJ/mol}}{4} = 116\ \text{kJ/mol}

This looks like an enormous stabilization — 20-fold larger than the surface case — but Step 4 shows this cannot be taken at face value.

Step 4 — Subtract the desolvation penalty. Burying a charged pair costs a large desolvation free energy, because both ions lose their favorable hydration shells (each ion's hydration free energy is roughly $-300$ to $-400$ kJ/mol for small monatomic ions, and tens to a hundred kJ/mol even for delocalized charges like guanidinium/carboxylate). Using the Born model for desolvation of a sphere of radius $a$ moving from water ( $\varepsilon_r=78.5$ ) to a cavity of $\varepsilon_r=4$ :

\Delta G_{\text{desolv}} = -\frac{e^2}{8\pi\varepsilon_0 a}\left(\frac{1}{\varepsilon_{\text{protein}}}-\frac{1}{\varepsilon_{\text{water}}}\right)

For $a \approx 0.20\ \text{nm}$ (typical ionic radius including first coordination), this penalty per ion is on the order of $150$ – $250\ \text{kJ/mol}$ , and it must be paid twice (once per ion). The net result is that most computational and experimental studies find buried, unpaired charges are strongly destabilizing, while a buried salt bridge — because the favorable pairing interaction partially offsets the desolvation cost — is only mildly stabilizing or even neutral, typically contributing a net few kJ/mol, similar in final magnitude to a surface salt bridge but for entirely different, cancelling reasons.

Limiting case check. As $\varepsilon_r \to \infty$ (a hypothetical infinitely polarizable medium), $E \to 0$ : any charge pair becomes invisible to itself, consistent with strongly polar solvents dissolving salts into fully independent, non-interacting ions — exactly what happens to NaCl in water at infinite dilution.

Go deeper · Derivation B · 3 minAdditivity and cancellation: why intramolecular hydrogen bonds barely stabilize folded proteins

Derivation BAdditivity and cancellation: why intramolecular hydrogen bonds barely stabilize folded proteins

Goal. Derive, from a simple thermodynamic cycle, why forming a backbone hydrogen bond during folding contributes far less free energy than the ~20 kJ/mol "bond strength" often quoted for an isolated N–H···O=C hydrogen bond.

Step 1 — Define the relevant thermodynamic cycle. Consider one backbone amide group that is hydrogen-bonded to bulk water in the unfolded state (donor N–H···OH₂, acceptor C=O···HOH) and hydrogen-bonded to a partner backbone carbonyl/amide in the folded state (an α-helical i, i+4 hydrogen bond, say). The relevant comparison is not

\Delta G_{\text{HB, naive}} = G(\text{no H-bond}) - G(\text{intramolecular H-bond})

which would overcount, but rather the cycle:

\Delta G_{\text{fold}} = \big[G(\text{H-bonded to partner}) - G(\text{H-bonded to water})\big]

because in the unfolded state the amide groups are never naked — they are always hydrogen-bonded to solvent water, which itself forms an extensive hydrogen-bond network.

Step 2 — Assign approximate bond strengths. Take the intramolecular backbone hydrogen bond (N–H···O=C in an α-helix) at roughly $E_{\text{intra}} \approx -25\ \text{kJ/mol}$ (a reasonable value for a good linear geometry, gas-phase-like estimate), and the corresponding hydrogen bond each partner would make to water, roughly $E_{\text{water}} \approx -21\ \text{kJ/mol}$ per water hydrogen bond (typical water–water or water–amide H-bond strength in bulk solvent, somewhat weaker due to competing thermal disorder and cooperative but also frustrated geometries in bulk water).

Step 3 — Compute the net enthalpic gain from the cycle. Since forming the intramolecular bond requires breaking two water hydrogen bonds (donor loses its H-bond to a water oxygen, acceptor loses its H-bond from a water hydrogen) and forming one intramolecular bond (which, being N–H···O=C, replaces both lost contacts with a single new bond — the geometry of an α-helical H-bond is one donor–one acceptor, so we compare bond-for-bond, not contact-for-contact):

\Delta H_{\text{net}} \approx E_{\text{intra}} - E_{\text{water}} = (-25) - (-21) = -4\ \text{kJ/mol}

This is the key result: the gross hydrogen bond energy is ~25 kJ/mol, but the net enthalpic contribution to folding is only about $-4\ \text{kJ/mol}$ per hydrogen bond, because almost all of the gross bond energy is already being "paid" by the water hydrogen bond that existed beforehand.

Step 4 — Include the entropic side of the ledger. Water molecules released from a specific, geometrically constrained hydrogen bond to an unfolded backbone group regain some translational and orientational freedom in bulk solvent, a small positive $\Delta S$ contribution favoring folding (a minor correction that this simplified treatment does not attempt to quantify precisely, but which pushes the same direction as $\Delta H_{\text{net}}$ ). Meanwhile, fixing the backbone dihedral angles into a single folded conformation costs substantial conformational entropy, roughly $-\,T\Delta S_{\text{conf}} \approx +8$ to $16\ \text{kJ/mol}$ per residue restricted to a single backbone conformation (a much larger term, discussed further under conformational entropy of folding).

Step 5 — Conclude on net hydrogen-bond stabilization. Combining, a single backbone hydrogen bond contributes on the order of only $-2$ to $-6\ \text{kJ/mol}$ net to protein stability — this is why single point mutations that remove one backbone hydrogen bond (e.g., proline substitutions that cannot donate, or engineered amide-to-ester "H-bond deletion" mutants) typically destabilize a protein by only 2–6 kJ/mol, not the naive 20–25 kJ/mol. Multiplied over dozens of backbone hydrogen bonds in a helix or sheet, however, this small per-bond term becomes a major contributor to overall stability (a protein with 30 helical H-bonds gains roughly $30 \times 4 = 120\ \text{kJ/mol}$ net, comparable to or exceeding the total folding free energy of 20–60 kJ/mol only because it is heavily opposed by the conformational entropy loss computed above — consistent with the observation that folded proteins are only marginally stable, by 20–60 kJ/mol out of hundreds of kJ/mol of gross interactions on each side).

Limiting case check. If a hydrogen-bond donor/acceptor pair were placed in a nonpolar solvent (no competing solvent hydrogen bonds, $E_{\text{water}}\to 0$ ), the net stabilization from forming the intramolecular bond would approach the full gross bond energy, $\Delta H_{\text{net}} \to E_{\text{intra}} \approx -25\ \text{kJ/mol}$ — consistent with the well-known experimental fact that hydrogen bonding contributes much more to stability of folded structures in nonpolar solvents or in the gas phase than in water, exactly the size and sign of effect this cycle predicts.

Worked examples

Example 1 (foundational): Screened electrostatic energy between a Lys⁺ side chain and a phosphate oxygen at the DNA backbone, contact distance 0.28 nm, in bulk water. Treat both as unit charges $e$ separated by $r = 2.8\times10^{-10}\ \text{m}$ , in water ( $\varepsilon_r = 78.5$ ):

E = \frac{e^2}{4\pi\varepsilon_0\varepsilon_r r} = \frac{(1.602\times10^{-19})^2}{4\pi(8.854\times10^{-12})(78.5)(2.8\times10^{-10})}

Numerator: $2.566\times10^{-38}\ \text{C}^2$ . Denominator: $4\pi(8.854\times10^{-12}) = 1.1127\times10^{-10}$ ; then $\times 78.5 = 8.734\times10^{-9}$ ; then $\times 2.8\times10^{-10} = 2.445\times10^{-18}$ .

E = \frac{2.566\times10^{-38}}{2.445\times10^{-18}} = 1.050\times10^{-20}\ \text{J per pair}

Multiply by $N_A$ :

E_{\text{molar}} = (1.050\times10^{-20})(6.022\times10^{23}) = 6.32\times10^{3}\ \text{J/mol} = 6.3\ \text{kJ/mol}

This single Lys–phosphate contact is worth about 6.3 kJ/mol — comparable to $1.5$ – $2.5\ RT$ at 298 K ( $RT = 2.48\ \text{kJ/mol}$ ) — modest on its own, but DNA-binding proteins typically make 8–12 such contacts simultaneously, so the additive electrostatic contribution to a transcription-factor–DNA complex, roughly $8\times6.3 \approx 50\ \text{kJ/mol}$ before counting desolvation and counterion release penalties, is a major term (though substantially offset by the entropically favorable release of condensed counterions from the DNA phosphate backbone upon protein binding, the polyelectrolyte effect, which is itself a separate, large, and favorable free energy term of similar or greater magnitude).

Example 2 (applied): Van der Waals contribution across a well-packed protein–ligand interface. A drug-like ligand buries roughly 40 heavy-atom pairs in van der Waals contact with a protein binding pocket upon binding (a typical value for a ~300–400 Da drug fully enclosed in a pocket). Assume each contact sits near its Lennard-Jones minimum with a typical well depth $\varepsilon \approx 0.4\ \text{kJ/mol}$ per pair (a representative average across C–C, C–N, C–O vdW contacts). The naive additive sum:

\Delta G_{\text{vdW,naive}} = 40 \times (-0.4\ \text{kJ/mol}) = -16\ \text{kJ/mol}

This is already a respectable free-energy contribution, roughly one-third to one-half of a typical drug's total binding free energy (a nanomolar inhibitor, $K_d \approx 10\ \text{nM}$ , corresponds to $\Delta G_{\text{bind}} = -RT\ln(1/K_d) = -(8.314\times10^{-3})(298)\ln(1\times10^8)$ ). Evaluating that reference number for comparison: $\ln(1\times10^8) = 18.42$ , so

\Delta G_{\text{bind}} = -(8.314\times10^{-3})(298)(18.42) = -45.6\ \text{kJ/mol}

So van der Waals contacts alone ( $-16\ \text{kJ/mol}$ ) supply roughly 35% of the total observed binding free energy of $-45.6\ \text{kJ/mol}$ for a nanomolar inhibitor — consistent with structure-based drug design heuristics that treat shape complementarity (maximizing vdW contact) as comparably important to specific polar contacts, and it is precisely why a single "buried hydroxyl with no H-bond partner" (a polar group that fails to find a partner and pays a desolvation penalty without compensation) is such a common and severe binding-affinity killer in medicinal chemistry: the vdW and hydrophobic terms are additive gains, but an unpaired polar group is a large uncompensated loss.

Example 3 (exam/competition level): Full additive decomposition and comparison with the observed $\Delta G$ for a designed protein–protein interface. A engineered protein–protein interaction has a measured dissociation constant $K_d = 50\ \text{nM}$ at 298 K. The interface buries 900 Å² of nonpolar surface area total (450 Å² from each partner) and forms 3 backbone/side-chain hydrogen bonds and 1 surface salt bridge, with an estimated 25 van der Waals atom-pairs near contact distance ( $\varepsilon \approx 0.4$ kJ/mol each). Estimate the total additive $\Delta G$ and compare to experiment, then find the implied conformational entropy penalty. Step A — convert $K_d$ to observed $\Delta G_{\text{bind}}$ .

\Delta G_{\text{obs}} = RT\ln K_d = (8.314\times10^{-3})(298)\ln(5.0\times10^{-8})

$\ln(5.0\times10^{-8}) = \ln(5.0) + \ln(10^{-8}) = 1.609 - 18.421 = -16.81$ .

\Delta G_{\text{obs}} = (2.478)(-16.81) = -41.65\ \text{kJ/mol}

Step B — hydrophobic term. Using $\gamma = 100\ \text{J mol}^{-1}\,\text{Å}^{-2}$ (a value near the upper end of literature estimates, appropriate for a well-packed interface at 298 K) and $\Delta A = 900\ \text{Å}^2$ :

\Delta G_{\text{hydrophobic}} = -(100\ \text{J mol}^{-1}\text{Å}^{-2})(900\ \text{Å}^2) = -90{,}000\ \text{J/mol} = -90.0\ \text{kJ/mol}

Step C — hydrogen bonds. Using the net per-bond stabilization derived above (roughly $-4$ kJ/mol net, since each intermolecular H-bond replaces a lost water H-bond just as in the intramolecular case):

\Delta G_{\text{HB}} = 3\times(-4\ \text{kJ/mol}) = -12\ \text{kJ/mol}

Step D — salt bridge. Using the surface (fully screened, $\varepsilon_r \approx 78.5$ ) value from Derivation 1:

\Delta G_{\text{ionic}} = -5.9\ \text{kJ/mol}

Step E — van der Waals.

\Delta G_{\text{vdW}} = 25\times(-0.4\ \text{kJ/mol}) = -10.0\ \text{kJ/mol}

Step F — sum the four favorable terms.

\Delta G_{\text{sum}} = -90.0 - 12.0 - 5.9 - 10.0 = -117.9\ \text{kJ/mol}

Step G — find the implied unfavorable term (conformational + translational/rotational entropy loss) by difference.

\Delta G_{\text{obs}} = \Delta G_{\text{sum}} - T\Delta S_{\text{unfavorable, as a positive } \Delta G_{\text{loss}}}

Rearranged:

\Delta G_{\text{loss}} = \Delta G_{\text{sum}} - \Delta G_{\text{obs}} = (-117.9) - (-41.65) = -76.3\ \text{kJ/mol lost, i.e. } +76.3\ \text{kJ/mol unfavorable}

This large positive residual — roughly 65% as large as the total favorable sum — is physically the cost of freezing out relative translational and rotational freedom of the two proteins upon complex formation (loss of 6 degrees of roto-translational freedom, typically estimated experimentally and computationally at $+40$ to $+60\ \text{kJ/mol}$ at standard states) plus residual side-chain conformational entropy loss at the interface ( $+10$ to $+30$ kJ/mol depending on interface size). The computed $+76.3\ \text{kJ/mol}$ falls squarely within this expected range, giving confidence that the additive decomposition, while approximate, captures the correct physics and correct order of magnitude for every term.

Limiting case check for Example 3. If the interface buried zero nonpolar area ( $\Delta A \to 0$ , an all-polar interface), $\Delta G_{\text{hydrophobic}}\to 0$ and the sum would drop to only $-27.9\ \text{kJ/mol}$ , likely insufficient to overcome even a modest entropy loss — correctly predicting that essentially all high-affinity (nanomolar) protein–protein interfaces observed in nature bury substantial hydrophobic surface, exactly as found in interface statistics from structural databases.

Limiting cases, scope & common traps

Force	Scales as	Typical magnitude (water)	Dominant regime	Key trap
Ionic	$1/r$ , $\propto 1/\varepsilon_r$	2–8 kJ/mol (surface); highly variable buried	Long range, additive at charged interfaces (DNA, membranes)	Quoting vacuum bond energies (100s of kJ/mol) as if realized in water
Hydrogen bond	steep ( $1/r^{2-4}$ ), $\cos^2\theta$	net 2–6 kJ/mol per bond in water	Directionality dictates secondary structure	Treating gross bond strength (~20 kJ/mol) as the folding contribution, ignoring the competing water H-bond
Van der Waals	$1/r^6$ attractive, $1/r^{12}$ repulsive	0.1–1 kJ/mol per atom pair	Additive over close-packed, shape-complementary interfaces	Dismissing as "negligible" without recognizing the multiplicative effect of thousands of contacts
Hydrophobic	$\propto \Delta A_{\text{nonpolar}}$ (area, not distance)	~25–100 J/(mol·Å²); tens of kJ/mol per typical interface	Dominant driving force for folding/binding at 298 K	Calling it an "attraction between oily groups" rather than a solvent entropy effect

Limiting cases to check every time:

As $\varepsilon_r \to \infty$ : all ionic interactions vanish — consistent with complete salt dissociation in a highly polar medium.
As $T \to 0$ : the hydrophobic effect's entropic driving force weakens and can even reverse sign (cold denaturation of proteins is a direct experimental manifestation — many proteins unfold on cooling toward 0 °C because the entropic penalty of ordering water around exposed nonpolar groups shrinks faster than the enthalpic terms).
At equal dielectric on both sides of a membrane or interface (no dielectric discontinuity), the desolvation penalty for burying a charge vanishes, and the buried-charge scenario in Derivation 1 simplifies back to the bulk-water case.
Setting the number of hydrogen bonds or vdW contacts to 1 recovers the single-interaction, per-bond energies used as building blocks throughout — a useful sanity check that additive sums reduce correctly to their single-term limit.

Common trapMisconception: "Hydrophobic groups attract each other"

The phrase "hydrophobic interaction" invites the false image of oily molecules being pulled toward each other by some intrinsic attractive force, analogous to van der Waals or ionic attraction. In fact, the direct van der Waals attraction between two methyl groups in water is unremarkable and nearly the same as it would be in any other medium. The actual driving force is what water does in their absence of contact: ordering itself around exposed nonpolar surface at an entropic cost. Removing that surface from water contact — regardless of whether the two nonpolar groups end up touching each other or simply both leave the water — is what releases the entropic penalty. This is why the hydrophobic effect scales with buried surface area, not with any pairwise interatomic distance, and why it is properly classified as a solvent reorganization phenomenon, not a direct intersolute force.

Common trapMisconception: bond energies are simply additive across a binding interface

It is tempting to add up textbook bond strengths — "3 hydrogen bonds at 20 kJ/mol plus a salt bridge at 20 kJ/mol equals 80 kJ/mol of binding energy." This overcounts dramatically because it ignores the reference state: in the unbound state, every one of those donor/acceptor and charged groups was already interacting favorably with water. The correct comparison, as shown in the derivations above, is always a difference between two solvated states, and the correctly computed net values (a few kJ/mol per hydrogen bond, a few kJ/mol per surface salt bridge) are five to ten times smaller than gross bond strengths. Only the hydrophobic and van der Waals terms survive this correction relatively undiminished, which is precisely why they dominate real binding free energies.

The chemist behind it

Origins · 1 minWalter Kauzmann and how this was discovered

Walter Kauzmann

1916–2009 · American chemist who formalized the hydrophobic effect

Walter Kauzmann, working at Princeton in the 1950s on the thermodynamics of protein denaturation, was the first to argue clearly and quantitatively that the dominant force stabilizing folded proteins was not hydrogen bonding (the prevailing assumption of the era, championed by Linus Pauling's structural work) but the transfer of nonpolar side chains out of water — an entropic phenomenon he analyzed by comparing the thermodynamics of hydrocarbon solubility in water to protein unfolding thermodynamics. His 1959 review, drawing an explicit quantitative parallel between the free energy, enthalpy, and entropy of transferring hydrocarbons from water to nonpolar solvents and the corresponding thermodynamics of protein denaturation, is the founding document of the modern "hydrophobic effect" as a distinct, quantifiable, and dominant thermodynamic force in biochemistry. Kauzmann's insight reoriented an entire generation of protein chemists away from a purely hydrogen-bond-centric view of folding and toward the entropy-based framework used throughout modern structural biology, directly anticipating the free-energy decompositions used in every contemporary computational and experimental analysis of binding and folding.

History

- 1936 — Linus Pauling and collaborators establish the electrostatic/quantum picture of the hydrogen bond and its role in ice and biomolecules.

- 1948 — Pauling proposes the α-helix and β-sheet, cementing hydrogen bonding as the presumed dominant force in protein structure.

- 1959 — Walter Kauzmann publishes his landmark analysis identifying the hydrophobic effect as the dominant driving force in protein folding.

- 1971 — Chothia and Lee introduce solvent-accessible surface area as the quantitative basis for the area-proportional hydrophobic free energy model still used today.

- 1980s–1990s — Systematic double-mutant cycle and alanine-scanning studies (Fersht, Wells, and others) provide direct experimental measurement of net hydrogen-bond and salt-bridge contributions to protein and interface stability, confirming the small net per-bond values derived from thermodynamic cycles.

Why is water's dielectric constant so much higher than that of a typical organic solvent?

Water's relative permittivity of ~78.5 arises because each water molecule carries a substantial permanent dipole moment (1.85 D) and, being small and freely tumbling in the liquid, can reorient rapidly to align with and screen an external or local electric field. Organic solvents like hexane have negligible permanent dipoles and only weakly polarizable electron clouds, giving $\varepsilon_r \approx 2$ ; more polar aprotic solvents like DMSO ( $\varepsilon_r \approx 47$ ) fall in between because their dipoles reorient but less freely and less strongly than water's extensively hydrogen-bonded network. This reorientation is fast (picosecond timescale) compared to most biological processes, so the dielectric screening described by Coulomb's law with $\varepsilon_r$ is an excellent approximation for equilibrium biochemical calculations, though it breaks down for very fast processes like electron transfer, where the solvent cannot fully reorient during the event itself.

If hydrogen bonds contribute so little net free energy, why do proteins care so much about satisfying every possible hydrogen bond donor and acceptor in the folded state?

Because the comparison is not symmetric. While a satisfied hydrogen bond (partner-to-partner) contributes only a small net stabilization relative to a bond to water, an unsatisfied hydrogen bond donor or acceptor buried in a nonpolar protein core — one with no partner at all — pays the full desolvation penalty of losing its water hydrogen bond without any compensating replacement. This asymmetry, sometimes called the "hydrogen bond satisfaction" or "buried polar group" penalty, is typically 15-25 kJ/mol unfavorable per unsatisfied group, far larger than the 2-6 kJ/mol gained by a properly satisfied one. This is why buried unpaired polar atoms are so destabilizing and why evolution and protein engineers alike work hard to ensure every buried polar group finds a partner.

Does the hydrophobic effect have different thermodynamic signatures at different temperatures?

Yes, and this is one of its most distinctive experimental fingerprints. At room temperature (~298 K), transferring a hydrocarbon from water to a nonpolar phase is entropy-driven: $\Delta H$ is often small or even slightly positive (unfavorable), while $-T\Delta S$ is large and negative (favorable), because ordered water around the nonpolar surface is released into bulk. At higher temperatures (~350-370 K), the entropic contribution shrinks (because bulk water itself becomes less structured, reducing the entropy difference to be gained) while the enthalpic contribution becomes more favorable, so the hydrophobic effect becomes progressively more enthalpy-driven. This is directly responsible for the characteristic large positive heat capacity change ( $\Delta C_p$ ) observed in protein unfolding calorimetry, itself considered the thermodynamic signature of hydrophobic burial.

Why do textbooks sometimes state hydrogen bond strengths of 20-40 kJ/mol while this article uses net values of only a few kJ/mol — is one of these wrong?

Neither is wrong; they describe different quantities. The 20-40 kJ/mol figures are gas-phase or vacuum bond dissociation energies (or gas-phase computational estimates), describing the intrinsic strength of the donor-acceptor interaction with no competing solvent. The 2-6 kJ/mol figures used throughout this article are net free energies of forming that bond in the specific context of aqueous biochemistry, where the alternative state is not "no bond" but "a bond to water of nearly comparable strength." Both numbers are correct for the question they answer; conflating them is the single most common error in introductory treatments of hydrogen bonding in biology.

How does ionic strength (added salt) affect noncovalent interactions in biochemistry?

Added salt ions in solution can screen electrostatic interactions further, beyond the dielectric screening already captured by $\varepsilon_r$ , through Debye-Hückel screening: mobile counterions cluster around any fixed charge and reduce its effective range and strength at physiological ionic strengths (~150 mM), described by a Debye screening length of roughly 0.8 nm at that ionic strength. This is why some salt bridges and electrostatic protein-protein or protein-DNA interactions weaken measurably as buffer salt concentration is raised experimentally (a standard biochemical assay for distinguishing electrostatic from hydrophobic or hydrogen-bonding contributions to binding), while purely hydrophobic and van der Waals contributions are comparatively insensitive to ionic strength, providing an experimental handle for decomposing binding free energy into its component forces.

Can van der Waals interactions ever be repulsive rather than attractive?

Yes, and this is essential to their role in molecular recognition. The Lennard-Jones potential has both an attractive $r^{-6}$ term (always present, from London dispersion) and a steep repulsive $r^{-12}$ term (from Pauli exclusion once electron clouds begin to overlap). At distances shorter than the potential minimum $r_{\min}=2^{1/6}\sigma$ , the net van der Waals interaction is repulsive and rises extremely steeply, which is precisely what gives atoms their effective "size" and enforces steric exclusion. This repulsive wall is what makes molecular shape complementarity possible at all — without it, molecules could interpenetrate freely and there would be no structural basis for specific binding.

Why is the hydrophobic effect considered the single largest contributor to protein folding free energy, ahead of hydrogen bonds and van der Waals contacts?

Because it acts on a large area rather than a discrete number of point contacts. A modestly sized globular protein buries several thousand square angstroms of nonpolar surface area upon folding, and even a conservative surface-tension coefficient of 25-50 J/(mol·Å²) yields tens of kilojoules per mole of favorable free energy from this term alone — comparable to or larger than the summed net contributions of all backbone hydrogen bonds and all van der Waals contacts combined, as demonstrated in the worked interface example above. This is precisely the insight that Walter Kauzmann formalized in 1959 and that reoriented the field away from a purely hydrogen-bond-centric view of protein stability.