OPLS-AA/1.14*CM1A Force Field Analysis

• OPLS-AA/1.14*CM1A is a fixed-charge small-molecule parametrization built on OPLS-AA that scales CM1A charges by 1.14 to enhance hydration free energy accuracy.
• It employs an additive energy decomposition with geometric combining rules and simulates glycosidic linkers, ring puckering, and conformer kinetics in aqueous sucrose.
• The LBCC variant locally redistributes charges to correct glucopyranose chair biases, leading to improved agreement with NMR observables and force-field accuracy.

Searching arXiv for papers on OPLS-AA/1.14*CM1A, LigParGen, and related force-field optimization. OPLS-AA/1.14*CM1A is a fixed-charge small-molecule parametrization built on OPLS-AA (Optimized Potentials for Liquid Simulations All-Atom) in which CM1A partial charges are scaled by 1.14. CM1A is derived from semiempirical AM1 with a bond-charge increment scheme, and the 1.14 scaling was introduced because it improves hydration free energies for general organic molecules. In the microsecond-scale study of sucrose in aqueous solution, the model was assessed together with the OPLS-AA/1.14*CM1A-LBCC variant and GLYCAM06, with emphasis on glycosidic linkage conformers, conformer lifetimes, glucopyranose and fructofuranose puckering, and agreement with NMR $J$-coupling constants and ultrasonic spectra (Deshchenya et al., 22 Sep 2025).

1. Force-field architecture and charge model

The OPLS-AA/1.14*CM1A model inherits the additive OPLS-AA energy decomposition. In a system-specific OPLS-AA optimization study, the force field is written as the sum of bond, angle, torsion, improper, van der Waals, and electrostatic terms, with pairwise Lennard-Jones parameters obtained from geometric combining rules, $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$ and $$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$, and with scaled 1–4 nonbonded interactions represented by a factor $f_{ij}$ (Hu et al., 2023).

Within that additive framework, OPLS-AA/1.14*CM1A replaces the charge assignment by using CM1A partial charges scaled by 1.14. In the sucrose study, the parameter source is Dodda et al., identified in the paper as LigParGen/OPLS-AA/1.14*CM1A. The same study also evaluates OPLS-AA/1.14*CM1A-LBCC, where LBCC denotes Localized Bond-Charge Correction. LBCC redistributes partial charges locally on specific bonds in order to correct systematic deficiencies in CM1A charges; relative to plain 1.14*CM1A, the reported change is local charge redistribution, and no torsion changes are reported in the paper (Deshchenya et al., 22 Sep 2025).

This distinction is central for carbohydrates. The sucrose results show that the difference between plain 1.14*CM1A and LBCC is not a minor implementation detail: the localized bond-charge redistribution changes the balance between ring puckering and glycosidic conformer stability. A common misconception is that density reproduction alone establishes a satisfactory carbohydrate model. The sucrose simulations show otherwise, because plain OPLS-AA/1.14*CM1A reproduces solution densities and diffusion in prior validation yet still exhibits an incorrect glucopyranose conformational preference.

2. Realization in aqueous sucrose simulations

The detailed characterization of OPLS-AA/1.14*CM1A for sucrose was carried out in explicit TIP4P/2005 water. The simulation protocol used OpenMM, with energy minimization by L-BFGS, NVT equilibration, NPT equilibration to determine equilibrium densities, and NVT production. Periodic boundary conditions were used; SHAKE was applied to all bonds involving H; the time step was 2 fs; Lennard-Jones and electrostatics were truncated at 12 Å; long-range corrections to pressure and energy were applied; and PME was used for Coulomb long-range interactions. The thermostat was Nosé–Hoover in NVT, and the barostat was Monte Carlo in NPT during equilibration (Deshchenya et al., 22 Sep 2025).

Three sucrose mass fractions were examined: $w_s = 20\%$, $30\%$, and $50\%$. Each box contained 200 sucrose molecules, with 15,200 water molecules at $20\%$, 8,867 at $30\%$, and 3,800 at $50\%$. Production trajectories were 1.0 $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$0s per system. For OPLS-AA/1.14*CM1A-LBCC, all three concentrations were simulated to 1 $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$1s; for GLYCAM06 and plain OPLS-AA/1.14*CM1A, $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$2 solutions were simulated to 1 $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$3s for comparison. Dihedrals $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$4 and $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$5 were recorded every 200 fs, and coordinates were saved every 5 ps (Deshchenya et al., 22 Sep 2025).

The equilibrium densities obtained in NPT were reported explicitly. For OPLS-AA/1.14*CM1A-LBCC they were $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$6 g/cm$\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$7 at $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$8, $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$9 g/cm$$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$0 at $$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$1, and $$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$2 g/cm$$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$3 at $$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$4. For GLYCAM06 at $$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$5, the reported value was $$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$6 g/cm$$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$7. The authors state that LBCC reproduces similar densities and diffusion to plain 1.14*CM1A and does not overaggregate sucrose (Deshchenya et al., 22 Sep 2025).

These details matter because the force-field comparison was not based on isolated conformers or dilute-limit calculations alone. It was carried out at finite concentrations and on microsecond trajectories, precisely where slow ring transitions and concentration-dependent conformer lifetimes become visible.

3. Glycosidic free-energy landscape and conformer kinetics

For sucrose, the glycosidic linkage dihedrals were defined as $$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$8 and $$\varepsilon_{ij} = \sqrt{\varepsilon_i \varepsilon_j}$$9. The free-energy surface was constructed as

$f_{ij}$0

where $f_{ij}$1 was computed by histogramming the trajectory over a $f_{ij}$2 grid and normalizing. Under OPLS-AA/1.14*CM1A-LBCC, three local minima, denoted M0, M1, and M2, were observed at all concentrations; the $f_{ij}$3 distribution was trimodal, whereas the $f_{ij}$4 distribution was unimodal (Deshchenya et al., 22 Sep 2025).

At $f_{ij}$5 sucrose, the reported LBCC minima were M0 at $f_{ij}$6 with $f_{ij}$7 kcal/mol, M1 at $f_{ij}$8 with $f_{ij}$9 kcal/mol, and M2 at $w_s = 20\%$0 with $w_s = 20\%$1 kcal/mol. The transition pathway M2 $w_s = 20\%$2 M0 $w_s = 20\%$3 M1 was identified, with free-energy maxima $w_s = 20\%$4 kcal/mol and $w_s = 20\%$5 kcal/mol along that path (Deshchenya et al., 22 Sep 2025).

Conformer lifetimes were extracted from an indicator autocorrelation analysis. If $w_s = 20\%$6 equals 1 when a molecule is in basin $w_s = 20\%$7 at time $w_s = 20\%$8 and 0 otherwise, the autocorrelation was defined as

$w_s = 20\%$9

fitted to

$30\%$0

with the lifetime

$30\%$1

For LBCC at 293 K, the reported lifetimes were: at $30\%$2, $30\%$3 ns, $30\%$4 ns, and $30\%$5 ns; at $30\%$6, $30\%$7 ns, $30\%$8 ns, and $30\%$9 ns; and at $50\%$0, $50\%$1 ns, $50\%$2 ns, and $50\%$3 ns. No 95% confidence intervals or transition rates were reported. M0 has the longest lifetimes, and M2 is longer lived than M1, consistent with the larger M2 $50\%$4 M0 barrier (Deshchenya et al., 22 Sep 2025).

The concentration dependence is explicit: lifetimes increase with concentration. The paper further reports that the weighted-average lifetime agrees with the disaccharide-specific ultrasonic relaxation time of approximately $50\%$5 ns at 298 K and 1 mol/L, corresponding to about $50\%$6 sucrose by mass. This places the force-field evaluation on both structural and dynamical grounds.

4. Ring puckering, chair inversion, and the $50\%$7 bias

The main deficiency of plain OPLS-AA/1.14*CM1A in the sucrose study is not a glycosidic torsion parameter reported in isolation, but a ring-puckering bias that propagates into the glycosidic conformer ensemble. Fructofuranose puckering was analyzed with Altona–Sundaralingam parameters using the phase $50\%$8, while glucopyranose puckering was analyzed with Cremer–Pople coordinates. For chair-state identification in glucopyranose, the study used the criterion $50\%$9 when $20\%$0 and $20\%$1 when $20\%$2; the explicit Cremer–Pople equations were not given in the paper (Deshchenya et al., 22 Sep 2025).

For fructofuranose, GLYCAM06 showed minima near $20\%$3 and $20\%$4. Plain OPLS-AA/1.14*CM1A showed three minima near $20\%$5, $20\%$6, and $20\%$7. OPLS-AA/1.14*CM1A-LBCC showed a global minimum near $20\%$8 and additional local minima around $20\%$9. The paper states that the solution-state flexibility mostly in the northern hemisphere is consistent with prior data for the OPLS force fields (Deshchenya et al., 22 Sep 2025).

The more consequential difference appears in glucopyranose. GLYCAM06 and OPLS-AA/1.14*CM1A-LBCC both predominantly sample the $30\%$0 chair, with reported populations of about $30\%$1–$30\%$2, and sample some $30\%$3 and non-chair conformations. For LBCC, metadynamics gives $30\%$4 kcal/mol and the global minimum favors $30\%$5. By contrast, plain OPLS-AA/1.14*CM1A shows a significant increase in the $30\%$6 population, and metadynamics places the global minimum near $30\%$7, meaning that $30\%$8 is the most stable form under that force field, in contradiction to experiment (Deshchenya et al., 22 Sep 2025).

This $30\%$9 bias stabilizes an additional glycosidic conformer, denoted M*, at approximately $50\%$0. The sucrose study concludes that the unconventional $50\%$1 glucopyranose chair stabilizes M*, and that M* is observed only with GLYCAM06 and plain OPLS-AA/1.14*CM1A in the full-population analysis. When Ramachandran plots were constructed using only molecules in the $50\%$2 chair, an M* minimum also appeared in the LBCC subset; however, because LBCC keeps the $50\%$3 population low, M* does not appear in the full LBCC ensemble (Deshchenya et al., 22 Sep 2025).

The significance is methodological as well as chemical. The result shows that an apparent glycosidic conformer discrepancy can arise from ring thermodynamics rather than from the glycosidic potential viewed in isolation. For sucrose, plain OPLS-AA/1.14*CM1A does not merely shift populations among M0, M1, and M2; it changes the qualitative conformer topology by stabilizing M* through an incorrect chair preference.

5. Experimental validation and comparative accuracy

The force fields were tested against NMR vicinal $50\%$4-coupling constants using Karplus-type relations. In the core form used in the paper,

$50\%$5

and agreement with experiment was quantified by

$50\%$6

The couplings were computed from 16 dihedrals using published Karplus-type equations; the main text notes also that extended Karplus forms with higher harmonics and electronegativity corrections were used for substituted systems, but the exact coefficients were not listed in the main text (Deshchenya et al., 22 Sep 2025).

The reported mean absolute errors show a clear ranking between plain OPLS-AA/1.14*CM1A and OPLS-AA/1.14*CM1A-LBCC.

Observable class	OPLS-AA/1.14*CM1A MAE (Hz)	OPLS-AA/1.14*CM1A-LBCC MAE (Hz)
Glucopyranose ring couplings	2.18	0.79
Fructofuranose ring couplings	0.98	0.74
Glycosidic pairs	1.47	0.76

GLYCAM06 provides intermediate performance in these comparisons, with reported MAEs of 0.97 Hz for glucopyranose ring couplings, 0.77 Hz for fructofuranose ring couplings, and 0.82 Hz for glycosidic pairs. The paper states that LBCC achieves the best agreement overall and especially for glycosidic pairs, whereas plain OPLS-AA/1.14*CM1A performs worst, driven by the incorrect glucopyranose conformational distribution and the associated $50\%$7 bias. Sample glycosidic couplings reinforce the point: for C2$50\%$8–C2$50\%$9, the simulated values are 2.5 Hz for GLYCAM06, 2.2 Hz for plain OPLS-AA/1.14*CM1A, and 2.0 Hz for LBCC against an experimental 2.0 Hz; for C2$\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$00–H1$\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$01, the simulated values are 4.5, 5.2, and 5.4 Hz against an experimental 3.9 Hz (Deshchenya et al., 22 Sep 2025).

Experimental validation was not limited to NMR. The weighted-average glycosidic lifetime from LBCC agrees with the disaccharide-specific ultrasonic relaxation time measured by Behrends and Kaatze, reported as approximately $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$02 ns at 298 K and 1 mol/L. On the solution-structure side, both LBCC and plain 1.14*CM1A with TIP4P/2005 were reported in prior validation to reproduce solution densities and diffusion and to avoid overaggregation, whereas GLYCAM06 at $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$03 sucrose formed a single large sucrose cluster and therefore strongly overaggregated sucrose despite yielding a comparable density at that concentration (Deshchenya et al., 22 Sep 2025).

The combined implication is narrow but important: a force field may reproduce density while still failing on ring populations and NMR observables. In the sucrose case, plain OPLS-AA/1.14*CM1A falls into exactly that category.

6. Suitability, limitations, and relation to broader OPLS-AA workflows

For microsecond sucrose dynamics in water, the reported conclusion is unambiguous. OPLS-AA/1.14*CM1A-LBCC with TIP4P/2005 is identified as the most suitable of the tested force fields because it captures the glycosidic conformer landscape M0, M1, and M2 without M* in the full ensemble, produces lifetimes consistent with ultrasonic relaxation, gives the best agreement with NMR $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$04-couplings, and does not overaggregate sucrose. Plain OPLS-AA/1.14*CM1A is reported to show a serious bias toward the $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$05 glucopyranose chair, to stabilize the nonphysical M* glycosidic conformer, and to degrade agreement with NMR; it is therefore not recommended for accurate sucrose conformational dynamics (Deshchenya et al., 22 Sep 2025).

The limitations of the sucrose assessment are also explicit. Chair inversion and ring conformational transitions occur on microsecond timescales, and convergence for ring populations was not achieved within 1 $\sigma_{ij} = \sqrt{\sigma_i \sigma_j}$06s for any force field. Enhanced sampling, specifically metadynamics, was therefore used to assess chair-inversion free energies. This suggests that even when microsecond trajectories are available, ring thermodynamics may remain a slow degree of freedom requiring dedicated free-energy methods (Deshchenya et al., 22 Sep 2025).

A broader OPLS-AA context is provided by a separate system-specific parameterization study. That work does not explicitly employ or recommend OPLS-AA/1.14*CM1A; instead, it optimizes OPLS-AA using QM-derived RESP, ESP, and Hirshfeld charges and extends the model with charge-transfer and polarization terms through CTPOL. It also states that the FFAFFURR workflow can accept any externally generated charge set, including CM1A with 1.14 scaling, as input for OPLS-AA parameterization, while cautioning that such choices should be validated against QM targets and molecular-dynamics observables (Hu et al., 2023).

That same study delineates the regime in which additive OPLS-AA models become insufficient. For neutral dipeptides and monovalent-cation systems, system-specific OPLS-AA optimization can approach low MAE values relative to QM energies, whereas for divalent cation–protein systems additive OPLS-AA, even after optimization, remains inadequate and requires explicit charge-transfer and polarization terms in CTPOL to improve both energetics and molecular-dynamics stability (Hu et al., 2023). A plausible implication is that OPLS-AA/1.14*CM1A should be viewed as a charge-model variant within an additive framework, not as a universal remedy for problems that arise from missing induction or charge-transfer physics.

In this narrower but well-characterized sense, OPLS-AA/1.14*CM1A is best understood through contrast with its LBCC correction. Plain 1.14*CM1A provides a useful baseline that can reproduce bulk solution properties in some cases, but in sucrose its local electrostatic balance is insufficient to preserve the experimentally favored glucopyranose chair distribution. The LBCC variant corrects that defect by modifying local charge distribution, thereby changing ring and glycosidic thermodynamics without any reported torsion change in the underlying parametrization.