From populations to absolute binding affinities in molecular simulations: exact volumetric terms and practical estimators
summary
The gist
From populations to absolute binding affinities in molecular simulations: exact volumetric terms and practical estimators presents a statistical-mechanics framework for computing equilibrium binding…
In short
The paper develops a statistical-mechanics framework to compute absolute binding constants (K) in molecular simulations by deriving exact volumetric terms and practical estimators. This method allows for unambiguous definition of thermodynamic state boundaries, making K directly computable from configuration histograms. It provides a reliable foundation for calculating binding affinities from biomolecular simulations.
Key concepts
- Canonical Ensemble
- This is a statistical ensemble used in molecular simulations that describes systems kept at constant temperature (T), volume (V), and number of particles (N). It forms the theoretical starting point for deriving the relationship between binding constants and statistical weights, allowing for rigorous thermodynamic calculations.
- Absolute Binding Free Energy ($\Delta F_0$)
- This is a central quantity used to determine binding affinity. The paper relates it to phase space integrals and statistical weights of bound versus unbound states. It is calculated using the relation $\Delta F_0 = \Delta F_{u\to b} - k_BT \log(V_{C0})$, where $V_{C0}$ represents the volume correction.
- Volumetric Correction ($V_{C0}$)
- This term accounts for the volume difference between the bound and unbound states, which is crucial for obtaining absolute binding constants. The framework precisely defines this volume, ensuring that simulations in finite boxes are correctly mapped to thermodynamic limits by accounting for the physical assumption of a homogeneous unbound state.
Terminology used across episodes
This episode discusses
- From populations to absolute binding affinities in molecular simulations: exact volumetric terms and practical estimators · Paper Radio
The paper
From populations to absolute binding affinities in molecular simulations: exact volumetric terms and practical estimators · Read on arXiv
Institute of Neuroscience and Medicine – Computational Biomedicine at Forschungszentrum Jülich GmbH · Department of Chemical Engineering at Virginia Tech
Biomolecular simulations of binding commonly confine the unbound ligand with volumetric restraints to accelerate convergence. However, a general account of how such restraints enter the volumetric term of the standard binding free energy, and of how this term should be evaluated from histogram data, is currently lacking. We take as our starting point the established relation, in the dilute limit, between the binding constant K and the relative populations of the bound and unbound states, which we re-derive from the canonical partition function of molecular solutes in explicit solvent. Building on this relation, we show that a restraint that leaves the bound state unaffected and confines a homogeneous unbound region modifies the volumetric term in a single way, by replacing the system volume with the volume of the region over which the unbound population is evaluated. This yields closed-form estimators for restrained simulations, computable from histograms of any suitably defined reaction coordinate, with the volumetric term kept explicit. Previous expressions based on the potential of mean force are recovered as a special case. The analysis also exposes two distinct errors in current practice: single-bin estimators, which reduce the bound state to one histogram bin and are not valid estimators of K, and working formulas that omit the bin width entering the volumetric term, which make the result depend on the unit of length. We apply the estimators to a Lennard-Jones dimer, the cucurbit[7]uril/1-adamantanol host--guest complex, and the galactonate--DgoT ligand--protein complex.
Transcript
Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.
Kai: We're looking at the paper: "From populations to absolute binding affinities in molecular simulations".
Mira: Here's the gist of the paper. From populations to absolute binding affinities in molecular simulations: exact volumetric terms and practical estimators presents a statistical-mechanics framework for computing equilibrium binding constants K in the dilute limit The resulting estimators are computable from histograms of any suitably defined reaction coordinate, and determine unambiguously how the boundaries of the thermodynamic states of interest must be defined This approach makes K directly computable, as constrained simulations can account for the volumetric contribution exactly, under the physically mild assumption of a homogeneous unbound state Theoretical Foundation The framework starts from the canonical ensemble and derives a central relation expressing K in terms of the statistical weights of the bound and unbound states The key result is Equation (10), which states that K = Pb Pu/V This equation relates the binding constant to the statistical weights of the bound and unbound states, where Pb and Pu are defined in terms of phase space integrals The general equation for absolute binding free energy is given by ∆F0 = ∆Fu→b −kBT log(VC0) Practical Estimators The paper provides practical estimators for both unrestrained and restrained simulations, including those that are unbiased or enhanced sampling simulations The statistical weight of a thermodynamic state s is approximated by Ps ≈ ∑ k∈s Pk, where the sum runs over the bins that define the state s For spherical symmetry, an estimator based on the pair correlation function g3(rk) is given by F(rk) = −kBT log g3(rk) −kBT log 4πCvr 2/∆r Application to Systems The formalism is applied to a hierarchy of systems, including a Lennard-Jones dimer, the cucurbit
7: uril/1-adamantanol host–guest complex, and the galactonate–DgoT ligand–protein complex In both molecular complexes, a single-bin estimate of the kind commonly employed departs from the theoretically correct one by ≈ 1 kcal/mol The derivation makes precise the volume entering the standard-state reference and the definition of the bound state, both of which are left implicit in earlier treatments The choice of these states thereby becomes a controlled input rather than a hidden assumption This approach offers a reliable foundation for current applications and a physically intuitive basis for future ones Tables 2 and 3 show what is at stake, as selecting the theoretically correct estimator shifts the binding free energy by roughly 1 kcal/mol The same reasoning applies to the unbound volume correction, which enters K through the term −kBT log(VC0) This matters in practice, since simulations are invariably performed in finite boxes far from the thermodynamic limit The construction is a starting point for developing new estimators, including for settings not treated here The reaction coordinate can be tailored to the system at hand, and the route to K is independent of how the configurations are generated The present treatment is a practical and unambiguous reference point for computing binding affinities from biomolecular simulations Supporting Information Details The supporting information details the derivation of practical estimators, including the hybrid estimator which combines an integrated bound state with a single-bin unbound reference The paper explicitly shows how the volumetric correction is obtained by directly substituting the restricted volume V R for V The final result in the thermodynamic limit is F = −kBT log
C ZS omega(N¯ u A,N¯ u B,N¯AB): , where N¯ u A, N¯ u B and N¯AB maximizes omega The generalisation of Eq.
Kai: (29) to the case of explicit solvent establishes a direct bridge between theory and explicit-solvent molecular simulations The equilibrium condition reduces to a point-wise identity, log zAB(S˜)zA(S˜)zB(S˜) =
B: /V ∀S˜ This is the generalisation of Eq.
Mira: (29) to the case of explicit solvent The resulting expression is equivalent to K = Pb Pu/V Simulation Details Molecular dynamics simulations were performed for the Lennard-Jones dimer and the cucurbit
7: uril/adamantanol complex using LAMMPS
70: and GROMACS 2022 patched with PLUMED 2.8
73: These simulations used well-tempered metadynamics
49: to sample the free energy profiles, and the analysis involved constructing probability histograms using kernel density estimation The results confirm that Estimates 1 and 2 agree with each other and with the estimate reported in Table 1 of the main text The final value for the binding affinity of the CB7/1-adamantanol complex using the extended-region estimator is-4.1362 kcal/mol The results show that Estimates 1 and 2 agree with each other and with the estimate reported in Table 1 of the main text The single-bin measure deviates by ≈ 1 kcal / mol The observed small magnitude of ∆F0 is physically consistent with the biological role of DgoT as a transporter rather than a tight binder for GAL The discrepancy with the predictions can be attributed to force-field inaccuracy The final estimate in Table 3 was obtained defining the bound state as the complementary of the unbound one shown in Fig.
Kai: 4, and it is discussed in the next section References
1: Michael K.
Paper summary: Kai: So, we're looking at this paper from arXiv called "From populations to absolute binding affinities in molecular simulations: exact volumetric terms and practical estimators". It sounds like they're trying to figure out how to calculate those absolute binding constants, K, in molecular simulations using a statistical mechanics framework.
Mira: Exactly. The main idea here is that they develop a general expression relating the binding constant K directly to the relative populations of the bound and unbound states, which makes things transparent. They claim this method explicitly shows where that volumetric term for the unbound state comes from, and it lets you track how an imposed volume restraint affects that expression.
Lev: So they're trying to make it computable even with simulations that impose constraints on volume, provided you assume the unbound state is homogeneous, which is a physically mild assumption. That sounds like something we need to check for real hardware setup.
Kai: Right, so the core thesis is about deriving a general expression for K based on these populations, and making sure the volume part of that equation is handled exactly when you use constrained simulations under those mild assumptions. But they aren't just stopping there; they are also providing practical estimators derived from histograms of any chosen reaction coordinate.
Mira: That's the other big claim—that these resulting estimators can be used to compute K unambiguously, and that you don't need some complicated setup to figure out how to define the boundaries of those thermodynamic states of interest. They say it makes K directly computable.
Lev: So what this means for us in terms of simulation overhead is that the computational cost shifts; you get an exact volumetric contribution when using restrained simulations, which is good if we can set up that restraint correctly and assume homogeneity. But they're also dealing with how to actually *get* those populations from the simulation data itself.
Kai: And they give practical estimators for both unrestrained and enhanced sampling simulations, including unbiased ones. For example, for spherical symmetry, they propose an estimator based on the pair correlation function g three of r k—that is F equals negative k Boltzmann temperature log g three of r k minus k Boltzmann temperature log four pi C volume to the power of two over the distance r squared divided by delta r.
Mira: That looks like a specific way to approximate that statistical weight, Ps, as the sum over all bins defining a state s. They're using that kind of functional form involving g three of r k and some geometric terms to get an estimate for the free energy contribution from the volume.
Paper summary: Lev: If we take that estimator and try to run it on real hardware with our existing metadynamics setups, we have to be careful about how we define those state bins because if you get it wrong, like in their example where a single-bin estimate deviates by about one kcal per mole, the whole binding affinity number gets off.
Kai: That discrepancy is important because they applied this formalism to systems like a Lennard-Jones dimer and even larger complexes, such as the cucurbit uril to adamantanol host guest complex, and the galactonate to dgoT ligand protein complex.
Mira: And what they found in those applications is that when you use a single-bin estimate, it doesn't match their theoretically correct one by about one kcal per mole in both molecular complexes. That shows that even with this new framework, the practical application still depends heavily on how you define those states and which estimator you choose.
Lev: So the point is that the choice of state definition becomes a controlled input instead of some hidden assumption made earlier in previous treatments. It’s about controlling the input so it’s not just some arbitrary choice in your simulation setup.
Kai: Yeah, that makes sense because they spend a lot of time clarifying exactly what volume enters the standard-state reference and how the bound state itself is defined in this new statistical mechanics approach. They are making those things explicit.
Mira: And they show that selecting the theoretically correct estimator shifts the binding free energy by roughly one kcal per mole, which is significant when you're trying to get accurate binding constants for biological systems. That means it matters a lot for what we measure in practice.
Lev: It also applies to the unbound volume correction term, which enters K through that term involving negative k Boltzmann temperature log V C zero over the volume V C zero. So even if you get the bound state right, you still have to be careful about how you handle that unbound reference volume correction when simulating in finite boxes far from the thermodynamic limit.
Kai: It sounds like this whole construction is a starting point for developing even newer estimators for settings they haven't covered yet, and crucially, the reaction coordinate you choose can be tailored to whatever system you are actually studying. The route to K isn't fixed by how you generate your configurations.
Paper summary: Mira: So what this implies for us in the long run is that this gives a reliable foundation for current applications we're doing now and provides a more physically intuitive basis for future work in molecular recognition simulations. It’s about getting the underlying physics right first.
Lev: We need to keep an eye on how researchers use these estimators, because the paper shows that even small differences in estimation can lead to noticeable shifts in the final affinity values you get out of your simulation runs. That's a concrete thing we have to account for when we move from theory into actual hardware experiments.
Kai: So, if you think about what this means for binding constants generally, it seems they’re moving towards a more rigorous way to handle the volume effects that were previously implicitly assumed in many methods. The paper "From populations to absolute binding affinities in molecular simulations: exact volumetric terms and practical estimators" gives us a clearer path to computing those constants from first principles.
Mira: It really does, because they’re breaking down exactly how the relative populations of the bound and unbound states translate into the actual binding constant K through these explicit volumetric terms. It makes that relationship much more transparent than what we used in earlier approaches.
Lev: From my side, it means when we look at scaling up to real hardware, like trying to run these simulations on systems with much larger particle numbers or complex restraints, we have a clearer understanding of where the computational complexity and the necessary assumptions lie.
Kai: The paper makes it clear that you can use histograms of any suitably defined reaction coordinate to get these estimators. That flexibility is what makes it useful across different types of molecular recognition problems, from small dimers to those big protein ligand complexes.
Mira: So we’re not just getting one magic formula; we're getting a statistical-mechanics framework that lets you choose your tools—your reaction coordinate and your estimator—and the theory tells you exactly how they connect to the actual binding constant K.
Lev: And remember, their limitation is that this framework assumes a physically mild assumption: that the unbound state is homogeneous. If your system isn't homogeneous, this whole exact volumetric accounting might break down, which is something we need to keep in mind when designing our next generation of error-correction methods.
Kai: Right, so the takeaway is that they’ve provided a more rigorous way to define and compute absolute binding affinities using statistical mechanics, making the volume contribution explicit and giving us better tools for simulation setup. That's where this paper sits for now.
Conclusion: Kai: So we're looking at this paper, "From populations to absolute binding affinities in molecular simulations: exact volumetric terms and practical estimators." It’s about getting those absolute binding constants, K, calculated from a statistical mechanics framework using things like population weights.
Mira: Yeah. The authors are showing how you can use those population numbers to get a direct link to the binding constant, which is usually hidden in the math. They focus on making the volume terms explicit so you know exactly what's contributing there.
Lev: From my end, it’s interesting that they’re focusing on these exact volumetric terms when we talk about real hardware. It sounds like they are trying to bridge the gap between how we simulate and what actually happens in a physical box.
Kai: Right. They present practical estimators that work for both unrestrained and restrained simulations, so you can see how those constraints affect the result directly.
Mira: The main point seems to be that this method lets you define the boundaries of your thermodynamic states clearly because it tells you exactly what inputs are needed for that calculation.
Lev: So, if I'm building an error correction scheme, I need to know if this framework is robust enough when we start imposing those volume restraints and assuming the unbound state is homogeneous.
Kai: Exactly. It’s a starting point for developing new estimators for things we haven't even simulated yet because you can tailor the reaction coordinate to whatever system you're working on.
Mira: And they show that by choosing the right estimator, you get a result that is physically consistent with what we expect from these biological systems.
Lev: It’s a good reference point for now, but I want to see if these estimators hold up when we move beyond simple molecular complexes and into more complex systems like explicit solvent simulations.
Kai: That’s the next step, I think. We'll talk more about how they generalize this theory to include explicit solvent in our next segment.
More episodes
- 2611.14082-Two atoms with strong light-induced dipolar interactions
- 2611.14040-Characterization of BNLs 20 and 30 micrometer thick LGADs for the Electron Ion Collider
- 2611.13399-Lift-element-based dynamic network of vortex gust-airfoil interactions
- 2611.13476-Hopfion Superlattices in 3D Twisted Moir'e Light Fields
- 2611.13642-Investigation of the timing performance of double-sided p-type 3D silicon sensors produced by CNM
- 2611.13830-SiPM optical readout of a PICOSEC Micromegas precise timing detector
- 2611.14152-Where are the shortest-path nodes in spatial networks?
- 2611.14219-Sub-Ice: semi-automated mapping of basal channels from ice shelf surface elevation data
- 2611.16341-SUMNO: Spatially Unrolled Multigrid Neural Operator for Fluid Flow Surrogates
- 2611.16090-Multi-UTE Myelin Fraction Mapping: Repeatability and Cross-Field-Strength Reproducibility at 1.5T and 3T