Spin-network-scaled MP2 (SNS-MP2) is a semi-empirical MP2-based method for computing the interaction energies of noncovalent complexes that is 6- to 7-fold more accurate than second-order Møller-Plesset perturbation theory (MP2). SNS-MP2 uses quantum chemical features of the complex under study in conjunction with a neural network to reweight terms appearing in the total MP2 interaction energy. SNS-MP2 has been fit to a training set of, and then benchmarked against a test set of, complete basis set (CBS)-extrapolated coupled-cluster singles, doubles, and perturbative triples (CCSD(T)) interaction energies, which are widely considered the gold standard for judging the accuracy of calculations of intermolecular interactions. The accuracy of SNS-MP2 is very close to the intrinsic accuracy of the CBS-extrapolated coupled-cluster methodology itself, and the method provides reliable per-conformation confidence intervals on the predicted interaction energies, a feature not available from any alternative method.
An open-source plugin for computing SNS-MP2 interaction energies and confidence intervals is available on GitHub without cost for both non-commercial and commercial use. The plugin is designed to be used with the Psi4 ab initio quantum chemistry software package.