Multitask Deep Ensemble Prediction of Molecular Energetics in Solution: From Quantum Mechanics to Experimental Properties

Abstract

The past few years have witnessed significant advances in developing machine learning methods for molecular energetics predictions, including calculated electronic energies with high-level quantum mechanical methods and experimental properties, such as solvation free energy and logP. Typically, task-specific machine learning models are developed for distinct prediction tasks. In this work, we present a multitask deep ensemble model, sPhysNet-MT-ens5, which can simultaneously and accurately predict electronic energies of molecules in gas, water, and octanol phases, as well as transfer free energies at both calculated and experimental levels. On the calculated data set Frag20-solv-678k, which is developed in this work and contains 678,916 molecular conformations, up to 20 heavy atoms, and their properties calculated at B3LYP/6-31G* level of theory with continuum solvent models, sPhysNet-MT-ens5 predicts density functional theory (DFT)-level electronic energies directly from force field-optimized geometry within chemical accuracy. On the experimental data sets, sPhysNet-MT-ens5 achieves state-of-the-art performances, which predict both experimental hydration free energy with a RMSE of 0.620 kcal/mol on the FreeSolv data set and experimental logP with a RMSE of 0.393 on the PHYSPROP data set. Furthermore, sPhysNet-MT-ens5 also provides a reasonable estimation of model uncertainty which shows correlations with prediction error. Finally, by analyzing the atomic contributions of its predictions, we find that the developed deep learning model is aware of the chemical environment of each atom by assigning reasonable atomic contributions consistent with our chemical knowledge.

Integrated Molecular Modeling and Machine Learning for Drug Design

“…In 2022, Xia et al developed a DL model named sPhysNet-MT-ens5 for predicting experimental hydration free energy and octanol/water partition coefficient (logP) using MMFF optimized geometry. Figure shows the overall workflow using the model for prediction.…”

“…Model performance comparison between sPhysNet-MT and other state-of-the-art models on the experimental hydration free energy and logP prediction. Results are adapted from the sPhysNet-MT manuscript …”

“…Motivated by the fact that many molecular modeling methods use the direct relationship between absolute energies and transfer energies to predict experimental properties with satisfactory results, − Xia et al developed a multitask DL model named sPhysNet-MT . This model simultaneously predicts electronic energies of molecules in gas, water, and octanol as well as the transfer free energies between these phases.…”

“…246,247 Multitask training on related tasks leads to more robust molecular representation learning. Motivated by the fact that many molecular modeling methods use the direct relationship between absolute energies and transfer energies to predict experimental properties with satisfactory results, 248−251 51 This model simultaneously predicts electronic energies of molecules in gas, water, and octanol as well as the transfer free energies between these phases. On their calculated data set, which includes molecules with energetics computed at the DFT-level, the sPhysNet-MT model achieves chemical accuracy on all tasks.…”

“…It is estimated that over 40% of drug candidates are withdrawn in preclinical tests due to ADME/T concerns . To address the problem, a number of computational models have been developed to predict the ADME/T properties of small molecules. − The accurate prediction of molecular properties requires models to learn more robust molecular representations for 1D SMILES, 2D graph, and 3D geometry …”