Retrospective on a decade of machine learning for chemical discovery

Abstract

Over the last decade, we have witnessed the emergence of ever more machine learning applications in all aspects of the chemical sciences. Here, we highlight specific achievements of machine learning models in the field of computational chemistry by considering selected studies of electronic structure, interatomic potentials, and chemical compound space in chronological order. Accurate solutions of the Schrödinger equation for the electrons in molecules and materials would vastly enhance our capability for chemical discovery, but computational cost makes this prohibitive. Since Dirac first exhorted us to find suitable approximations to bypass this cost 1 , much progress has been made, but much remains out of reach for the foreseeable future. The central promise of machine learning (ML) is that, by exploiting statistical learning of the properties of a few cases, we might leapfrog over the worst bottlenecks in this process. As visible from the publication record in the field (Fig. 1), over the decade since Nature Communications first appeared, machine learning has gained increasing traction in the hard sciences 2 , and has found many applications in atomistic simulation sciences 3. Here, we focus on the progress achieved in the last decade on three interrelated topics (i) electronic structure theory, broadly defined, (ii) universal force field models, as used for vibrational analysis or molecular dynamics applications, and (iii) first principles-based approaches enabling the exploration of chemical compound space. Basic challenges The central challenge of Schrödinger space is to use supervised learning from examples to find patterns that either accelerate or improve upon the existing human algorithms behind these technologies. In density functional theory (KS-DFT), this most often means improved approximate functionals; in quantum Monte Carlo (QMC), this is faster ways to find variational wavefunctions; in ab initio quantum chemistry such as coupled cluster considering single, double, and perturbative triple excitations (CCSD(T)), this is learned predictions of wavefunction amplitudes instead of recalculation for every system. In the condensed phase, molecular dynamics simulations yield a vast amount of useful thermodynamic and kinetic properties. Classical force fields cost little to run, but are often accurate only around the equilibrium. The only first-principles alternative is Kohn-Sham density functional theory (DFT), but its computational cost vastly reduces what is practical. A central challenge of configuration space is therefore to produce energies and forces from a classical

Evaluation of Thermochemical Machine Learning for Potential Energy Curves and Geometry Optimization

“…5 The number of molecules predicted to have a local or global minima after 2Å. 6 BAND-NN regularly would not predict energies for geometries with a bond stretch of 2Å or greater.…”

Big data and machine learning for materials science

“…Two converging movements are responsible for the revolution. The first may be referred to as “data-intensive discovery” [ 1 ], “e-Science”, or “big data” [ 3 , 4 ], a movement wherein massive amounts of data are transformed into knowledge. This is attained with various computational methods, which increasingly engage ML techniques, in a movement characterized by a transition in which data move from a “passive” to an “active” role.…”

Predicting phosphorescence energies and inferring wavefunction localization with machine learning

“…Machine learning (ML) has newfound relevance in quantum chemistry for accelerating simulations and providing predictions of ab initio quality. [14][15][16] Here, we leverage ML techniques and build an accurate and extensible ML model for phosphorescence energy that is enabled by its ability to account for electron localization associated with the spin transition.…”