Thermodynamics is fundamental for understanding design principles of natural metabolic processes and for engineering new efficient metabolic pathways. Accurate standard Gibbs reaction energies for biochemical reactions are of crucial importance in this regard. As accurate experimental Gibbs reaction energies exist only for a small fraction of metabolic reactions, research relies on empirical methods for the estimation of Gibbs reaction energies. Conventionally, group contribution methods are used to obtain such an estimate from the standard Gibbs formation energies of reactants and products obtained from functional group contributions.

Researchers from Harvard University, the University of Toulouse, Princeton University and the Weizmann Institute of Science have presented the first non-empirical high-throughput method for estimating standard Gibbs reaction energies for metabolic reaction from quantum chemistry. Using first-principle methods for the prediction of Gibbs reaction energies, or thermodynamic parameters in general, has several crucial advantages. First, first-principle methods are not limited by available experimental data. Moreover, they can take advantage of the hierarchy of quantum-chemical methods, allowing a defined trade-off between speed and accuracy.

The proposed method is a bottom-up method relying on an exhaustive sampling of conformers and protonation states of metabolites in solution. Each metabolite is represented by an ensemble of protonation states and each protonation state is represented by an ensemble of conformers. Short-range solvation effects are accounted for through explicit inclusion of a fixed number of water molecules (five or ten). Long-range electrostatic interactions are described by means of a continuum solvation model.

In their proof of principle study, the researchers tested their method on two test sets. The first test set allows comparison to accurate experimental standard Gibbs free reaction energie and comprises nine metabolic reactions. The second, broader test set comprises 113 metabolic reaction from the NIST-TECR database, providing a large-scale benchmark study for the proposed method. The second test set comprises nearly 6000 quantum-chemical calculations with a median run time for geometry optimization and harmonic analysis of a single conformer of 3.4 hours (16 CPUs).

The presented first-principle method provides an accuracy comparable to the accuracy of group contribution methods - if reactions including multiply charged anions are excluded. These multiply charged anions might require a better description of the solvent, both through a larger solvent shell and quantum-chemical methods that possess the correct asymptotic behavior of the potential.

The researchers propose several options to pursue on this promising path. One of them is replacing the density function calculations using B3LYP with long-range corrected functionals or switching directly to the linear-scaling DLPNO-CCSD(T) method.

We are eager to see where these will lead us on the way to accurate metabolic thermochemical methods.