Fast and accurate modeling of molecular atomization energies with machine learning.

By Matthias Rupp , Alexandre Tkatchenko, Klaus‐Robert Müller , O. A. von Lilienfeld. 2011

1,758 citations recorded by the source corpus.

We introduce a machine learning model to predict atomization energies of a diverse set of organic molecules, based on nuclear charges and atomic positions only. The problem of solving the molecular Schrödinger equation is mapped onto a nonlinear statistical regression problem of reduced complexity. Regression models are trained on and compared to atomization energies computed with hybrid density-functional theory. Cross validation over more than seven thousand organic molecules yields a mean absolute error of ∼10 kcal/mol. Applicability is demonstrated for the prediction of molecular atomization potential energy curves.

Subfields: Atomic and Molecular Physics, and Optics, Computational Theory and Mathematics, Materials Chemistry, Computational Chemistry & Drug Discovery

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Atomic and Molecular Physics, and Optics

In Computational Theory and Mathematics

In Materials Chemistry

In Computational Chemistry & Drug Discovery

Citations in the corpus

Theory-Guided Data Science: A New Paradigm for Scientific Discovery from Data (2016, background): “, scaling, translation, and rotational invariance in turbulence modeling [78], and symmetry invariance in computational chemistry [26], [79].”
Learning atoms for materials discovery (2018, background): “The input or feature of materials involves descriptors of constituents: Certain physical or chemical attributes of atoms are taken, depending on the materials property under prediction (11, 14, 17).”
Learning Graph-Level Representation for Drug Discovery (2017, methodology): “A wide-used conventional proposal is to utilize hand-crafted feature like ECFP(Rogers and Hahn 2010), Coulomb Matrix(Rupp et al. 2012) as input and feed it into conventional classifier like random forest and multilayer perceptron.”
SchNet: A continuous-filter convolutional neural network for modeling quantum interactions (2017, background): “In recent years, there have been increased efforts to use machine learning for the accelerated discovery of molecules and materials with desired properties [1–9].”
Automatic Chemical Design Using a Data-Driven Continuous Representation of Molecules (2016, neutral): “Third, a data-driven representation can leverage large sets of unlabeled chemical compounds to automatically build an even larger implicit library, and then use the smaller set of labeled examples to build a regression model from the continuous representation to the desired properties.”
Relevance of Rotationally Equivariant Convolutions for Predicting Molecular Properties (2020, background): “1 Related Work Molecular properties, which depend only on the atomic distance graph, are commonly predicted by kernel methods or Gaussian process regression [14, 15, 16, 17] or graph neural networks [18, 19], where ENNs are usually employed for predicting physical properties, which depend on atomic ”
Generalizing Convolutional Neural Networks for Equivariance to Lie Groups on Arbitrary Continuous Data (2020, background): “We then apply T( 3 )- equivariant LieConv layers to the full range of tasks in the QM9 dataset and report the results in Table 2.”
Graph Networks as a Universal Machine Learning Framework for Molecules and Crystals (2018, background): “se the ˚s to be multi-layer perceptrons with two hidden layers (Equation 7), given their ability to be universal approximators for non-linear functions.40 ˚(x) = W 3((W 2((W 1x+ b 1)) + b 2)) + b 3 (7) where is the modied softplus function10 acting as nonlinear activator, Ws are the kernel weights a”
Clebsch-Gordan Nets: a Fully Fourier Space Spherical Convolutional Neural Network (2018, methodology): “Atomization Energy Prediction Next, we apply our framework to the QM7 dataset [30, 31], where the goal is to regress over atomization energies of molecules given atomic positions ( p i ) and charges ( z i ).”
Neural Message Passing with Edge Updates for Predicting Properties of Molecules and Materials (2018, methodology): “…machine learning methods for predicting properties of molecules and materials have been based on hand crafted feature descriptors, such as Coulomb matrix (Rupp et al., 2012), bag-of-bonds (Hansen et al., 2015), ﬁngerprints (e.g. Rogers and Hahn 2010) and histogram of angles (Faber et al., 2017).”
Spherical CNNs (2018, methodology): “To this end Rupp et al. (2012) propose a distance measure between Coulomb matrices used within Gaussian kernels whereas Montavon et al. (2012) propose sorting C or random sampling index permutations.”