Contributor info: Prof. Jan H. Jensen

Maria H. Rasmussen, Diana S. Christensen, Jan H. Jensen

SciPost Chem. 2, 002 (2023) · published 27 June 2023 | Toggle abstract · pdf

While there is a great deal of interest in methods aimed at explaining machine learning predictions of chemical properties, it is difficult to quantitatively benchmark such methods, especially for regression tasks. We show that the Crippen logP model (J. Chem. Inf. Comput. Sci. 1999, 39, 868) provides an excellent benchmark for atomic attribution/heatmap approaches, especially if the ground truth heatmaps can be adjusted to reflect the molecular representation. The "atom attribution from finger prints"-method developed by Riniker and Landrum (J. Chem. Inf. Comput. Sci. 2013, 5, 43) gives atomic attribution heatmaps that are in reasonable agreement with the atomic contribution heatmaps of the Crippen logP model for most molecules, with average heatmap overlaps of up to 0.54. The agreement is increased significantly (to 0.75) when the atomic contributions are adjusted to match the fact that the molecular representation is fragment-based rather than atom-based (the finger print-adapted (FPA) ground truth vector). Most heatmaps and the corresponding FPA overlaps are relatively insensitive to the training set size and the results are close to converged for a training set size of 1000 molecules, although for molecules with low overlap some heatmaps change significantly. Using the "remove an atom" approach for graph convolutional neural networks (GCNNs) suggested by Matveieva and Polishchuk (J. Cheminform. 2021, 13, 41) we find an average heatmap overlap of 0.47 for the atomic contribution heatmaps of the Crippen logP model. Like the simpler attribution benchmarks for classification tasks that have come before it, this work sets the bar for regression tasks.

Mads Koerstz, Maria H. Rasmussen, Jan H. Jensen

SciPost Chem. 1, 003 (2021) · published 1 October 2021 | Toggle abstract · pdf

We show how fast semiempirical QM methods can be used to significantly decrease the CPU requirements for automated reaction mechanism discovery, using two different method for generating reaction products: graph-based systematic enumeration of all possible products and the meta-dynamics approach by Grimme (J. Chem. Theory. Comput. 2019, 15, 2847). We test the two approaches on the low-barrier reactions of 3-hydroperoxypropanal, which have been studied by a large variety of reaction discovery approaches and therefore provides a good benchmark. By using PM3 and GFN2-xTB for reaction energy and barrier screening the systematic approach identifies 64 reactions (out of 27,577 possible reactions) for DFT refinement, which in turn identifies the three reactions with lowest barriers plus a previously undiscovered reaction. With optimised hyperparameters meta-dynamics followed by PM3/GFN2-xTB-based screening identifies 15 reactions for DFT refinement, which in turn identifies the three reactions with lowest barrier. The number of DFT refinements can be further reduced to as little as six for both approaches by first verifying the transition states with GFN1-xTB. The main conclusion is that the semiempirical methods are accurate and fast enough to automatically identify promising candidates for DFT refinement for the low barrier reactions of 3-hydroperoxypropanal in about 15-30 minutes using relatively modest computational resources.