Representations and descriptors unifying the study of molecular and bulk systems

Rossi, Kevin; Cumby, James

doi:10.1002/qua.26151

Cited by 11 publications

(15 citation statements)

References 57 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The exhaustive approach of MBTR descriptor to documenting molecular features has led to very good predictive accuracy in machine learning of molecular properties (Stuke et al, 2019;Langer et al, 2020;Rossi and Cumby, 2020;Himanen et al, 2020) and this work is no exception. The lightweight CM descriptor does not perform nearly as well, but these two representations from physical sciences provide us with an upper and lower limit on predictive accuracy.…”

Section: Resultsmentioning

confidence: 65%

“…We transform the molecular structures in Wang's dataset into atomistic descriptors more suitable for machine learning than the atomic coordinates or the commonly used simplified molecular-input line-entry system (SMILES) strings. Optimal descriptor choices have been the subject of increased research in recent years (Langer et al, 2020;Rossi and Cumby, 2020;Himanen et al, 2020). We here test several descriptor choices: the many body tensor representation (Huo and Rupp, 2017), the Coulomb matrix (Rupp et al, 2012), the Molecular ACCess System (MACCS) structural key (Durant et al, 2002), a topological fingerprint developed by RDkit (Landrum et al, 2006) based on the daylight fingerprint (James et al, 1995) and the Morgan fingerprint (Morgan, 1965).…”

mentioning

confidence: 99%

See 1 more Smart Citation

Predicting Gas-Particle Partitioning Coefficients of Atmospheric Molecules with Machine Learning

Lumiaro¹,

Todorović²,

Kurtén³

et al. 2021

Preprint

View full text Add to dashboard Cite

Abstract. The formation, properties and lifetime of secondary organic aerosols in the atmosphere are largely determined by gas-particle partitioning coefficients of the participating organic vapours. Since these coefficients are often difficult to measure and to compute, we developed a machine learning model to predict them given molecular structure as input. Our data-driven approach is based on the dataset by Wang et al. (Atmos. Chem. Phys., 17, 7529 (2017)), who computed the partitioning coefficients and saturation vapour pressures of 3414 atmospheric oxidation products from the master chemical mechanism using the COSMOtherm program. We trained a kernel ridge regression (KRR) machine learning model on the saturation vapour pressure (Psat), and on two equilibrium partitioning coefficients: between a water-insoluble organic matter phase and the gas phase (KWIOM/G), and between an infinitely dilute solution with pure water and the gas phase (KW/G). For the input representation of the atomic structure of each organic molecule to the machine, we tested different descriptors. We find that the many-body tensor representation (MBTR) works best for our application, but the topological fingerprint (TopFP) approach is almost as good, and is significantly more cost effective. Our best machine learning model (KRR with a Gaussian kernel + MBTR) predicts Psat and KWIOM/G to within 0.3 logarithmic units and KW/G to within 0.4 logarithmic units of the original COSMOtherm calculations. This is equal or better than the typical accuracy of COSMOtherm predictions compared to experimental data (where available). We then applied our machine learning model to a dataset of 35,383 molecules that we generated based on a carbon 10 backbone functionalized with 0 to 6 carboxyl, carbonyl or hydroxyl groups to evaluate its performance for polyfunctional compounds with potentially low Psat. The resulting saturation vapor pressure and partitioning coefficient distributions were physico-chemically reasonable, and the volatility predictions for the most highly oxidized compounds were in qualitative agreement with experimentally inferred volatilities of atmospheric oxidation products with similar elemental composition.

show abstract

Section: Resultsmentioning

confidence: 65%

mentioning

confidence: 99%

Predicting Gas-Particle Partitioning Coefficients of Atmospheric Molecules with Machine Learning

Lumiaro¹,

Todorović²,

Kurtén³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…More specifically, it hinges on how well each format encapsulates the structural features relevant to the atmospheric behaviour. The exhaustive approach of the MBTR descriptor to documenting molecular features has led to very good predictive accuracy in machine learning of molecular properties (Stuke et al, 2019;Langer et al, 2020;Rossi and Cumby, 2020;Himanen et al, 2020), and this work is no exception. The lightweight CM descriptor does not perform nearly as well, but these two representations from physical sciences provide us with an upper and lower limit on predictive accuracy.…”

Section: Discussionmentioning

confidence: 68%

“…In KRR, unlike ridge regression, a nonlinear kernel is applied. This maps the molecular structure to our target properties in a high-dimensional space (Stuke et al, 2019;Rupp, 2015).…”

Section: Machine Learning Methods 231 Kernel Ridge Regressionmentioning

confidence: 99%

“…We transform the molecular structures in Wang's dataset into atomistic descriptors more suitable for machine learning than the atomic coordinates or the commonly used simplified molecular-input line-entry system (SMILES) strings. Optimal descriptor choices have been the subject of increased research in recent years (Langer et al, 2020;Rossi and Cumby, 2020;Himanen et al, 2020). We test several descriptor choices here: the many-body tensor representation (Huo and Rupp, 2017), the Coulomb matrix (Rupp et al, 2012), the Molecular ACCess System (MACCS) structural key (Durant et al, 2002), a topological fingerprint developed by RDKit (Landrum, 2006) based on the daylight fingerprint (James et al, 1995), and the Morgan fingerprint (Morgan, 1965).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Predicting gas–particle partitioning coefficients of atmospheric molecules with machine learning

et al. 2021

View full text Add to dashboard Cite

Abstract. The formation, properties, and lifetime of secondary organic aerosols in the atmosphere are largely determined by gas–particle partitioning coefficients of the participating organic vapours. Since these coefficients are often difficult to measure and to compute, we developed a machine learning model to predict them given molecular structure as input. Our data-driven approach is based on the dataset by Wang et al. (2017), who computed the partitioning coefficients and saturation vapour pressures of 3414 atmospheric oxidation products from the Master Chemical Mechanism using the COSMOtherm programme. We trained a kernel ridge regression (KRR) machine learning model on the saturation vapour pressure (Psat) and on two equilibrium partitioning coefficients: between a water-insoluble organic matter phase and the gas phase (KWIOM/G) and between an infinitely dilute solution with pure water and the gas phase (KW/G). For the input representation of the atomic structure of each organic molecule to the machine, we tested different descriptors. We find that the many-body tensor representation (MBTR) works best for our application, but the topological fingerprint (TopFP) approach is almost as good and computationally cheaper to evaluate. Our best machine learning model (KRR with a Gaussian kernel + MBTR) predicts Psat and KWIOM/G to within 0.3 logarithmic units and KW/G to within 0.4 logarithmic units of the original COSMOtherm calculations. This is equal to or better than the typical accuracy of COSMOtherm predictions compared to experimental data (where available). We then applied our machine learning model to a dataset of 35 383 molecules that we generated based on a carbon-10 backbone functionalized with zero to six carboxyl, carbonyl, or hydroxyl groups to evaluate its performance for polyfunctional compounds with potentially low Psat. The resulting saturation vapour pressure and partitioning coefficient distributions were physico-chemically reasonable, for example, in terms of the average effects of the addition of single functional groups. The volatility predictions for the most highly oxidized compounds were in qualitative agreement with experimentally inferred volatilities of, for example, α-pinene oxidation products with as yet unknown structures but similar elemental compositions.

show abstract

Accelerating the theoretical study of Li‐polysulfide adsorption on single‐atom catalysts via machine learning approaches

Andritsos

Rossi

2022

Int J of Quantum Chemistry

Self Cite

View full text Add to dashboard Cite

Li-S batteries are a promising alternative to Li-ion batteries, offering large energy storage capacity and wide operating temperature range. However, their performance is heavily affected by the Li-polysulfide (LiPS) shuttling. Computational screening of LiPS adsorption on single-atom catalyst (SAC) substrates is of great aid to the design of Li-S batteries which are robust against the LiPS shuttling from the cathode to the anode and the electrolyte. To facilitate this process, we develop a machine learning (ML) protocol to accelerate the systematic mapping of dominant local energy minima found with calculations based on the density functional theory (DFT), and, in turn, fast screening of LiPS adsorption properties on SACs. We first validate the approach by probing the potential energy surface for LiPS adsorbed on graphene decorated with a Fe-N 4 -C SAC. We identify minima whose binding energies are better or on par with the one previously reported in the literature. We then move to analyze the adsorption trends on Zn-N 4 -C SAC and observe similar adsorption strength and behavior with the Fe-N 4 -C SAC, highlighting the good predictive power of our protocol. Our approach offers a comprehensive and computationally efficient alternative to conventional approaches studying LiPS adsorption.

show abstract

Representations and descriptors unifying the study of molecular and bulk systems

Cited by 11 publications

References 57 publications

Predicting Gas-Particle Partitioning Coefficients of Atmospheric Molecules with Machine Learning

Predicting Gas-Particle Partitioning Coefficients of Atmospheric Molecules with Machine Learning

Predicting gas–particle partitioning coefficients of atmospheric molecules with machine learning

Accelerating the theoretical study of Li‐polysulfide adsorption on single‐atom catalysts via machine learning approaches

Contact Info

Product

Resources

About