COSMO-RS predictions of logP in the SAMPL7 blind challenge

Warnau, Judith; Wichmann, Karin; Reinisch, Jens

doi:10.1007/s10822-021-00395-5

Cited by 14 publications

(14 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Comparing our SMD q -DFT sp results with others obtained by similar methods, we note that in the SAMPL6 challenge, log P blind predictions for 11 small molecules rendered RMSE values between 0.44 and 1.29 log P units for quantum chemistry methods along with continuum solvent models − before the inclusion of empirical corrections . Furthermore, results for log P blind predictions for 22 small molecules from the SAMPL7 challenge included RMSE = 1.03 log P units for IEFPCM calculations and RMSE = 0.78 log P units for COSMO-RS calculations as the top performers for physical models. Again, this comparison demonstrates that the SMD q -DFT sp results provide competitive estimates of log P .…”

Section: Resultssupporting

confidence: 60%

Density Functional Theory with Implicit Solvents for Accurate Estimation of Aqueous and Organic Solvation Free Energies of Lignin Fragments

Colombari

Nascimento

Liu

et al. 2022

ACS Sustainable Chem. Eng.

View full text Add to dashboard Cite

Lignin is the main potential source of renewable aromatic chemicals, and solvation free energy (ΔG solv ) is a fundamental thermodynamic parameter for a compound in solution. Here, density functional theory with the electron density-based continuum solvent model (SMD-DFT) is employed to calculate ΔG solv for 283 neutral lignin depolymerization fragments and analogues (LDF&A) in water and organic solvents (1-octanol, cyclohexane, methyl-isobutyl ketone, and butyl acetate). Calculations for LDF&A are challenging due to the balance of hydrophilic oxygenations and hydrophobic aromatic rings with π orbitals. After correction of atomic van der Waals radii, calculated ΔG solv in water for a subset of 18 aromatic monomers is shown to outperform the accuracy of the FreeSolv benchmark. For organic solvents, comparisons through partition coefficients (log P) from several public datasets reveal that SMD-DFT calculations can accurately predict log P after accounting for systematic biases due to solvent admixtures not considered by the calculations. The presented results for LDF&A demonstrate that SMD-DFT provides useful estimates for log P and unprecedented accuracy for aqueous ΔG solv . The obtained datasets of aqueous and organic ΔG solv for LDF&A provide a key advancement toward accurate computer modeling of lignin processing into renewable fuels and chemicals.

show abstract

Section: Resultssupporting

confidence: 60%

Density Functional Theory with Implicit Solvents for Accurate Estimation of Aqueous and Organic Solvation Free Energies of Lignin Fragments

Colombari

Nascimento

Liu

et al. 2022

ACS Sustainable Chem. Eng.

View full text Add to dashboard Cite

show abstract

“…The PHYSPROP , data set contains around 14,500 molecules and their experimental logP values which are further curated following an established protocol , including removal of inorganics and mixtures, structural conversion and cleaning, normalization of specific chemotypes, removal of duplicates and manual inspection. The experimental octanol to water transfer free energy is calculated from experimental logP by the following thermodynamic equation Δ E o c t → w a t e r = R T ln ( 10 ) × log P where R is the ideal gas constant, and T is the temperature assumed to be 298.15 K in this work.…”

Section: Data Setmentioning

confidence: 99%

“…The experimental octanol to water transfer free energy is calculated from experimental logP by the following thermodynamic equation. 47…”

Section: ■ Data Setmentioning

confidence: 99%

“…In this work, motivated by traditional molecular modeling methods which utilize the direct relationship between absolute energies and transfer energies, − we present sPhysNet-MT, a multitask DL model which predicts molecular electronic energies in gas, water, and octanol phases directly and corresponding transfer free energies indirectly, from MMFF-optimized molecule geometries. Unlike some previous multitask models for less correlated tasks that usually perform worse than their single-task variants, , well-correlated tasks facilitate the model for simultaneous predictions, , and the tasks in sPhysNet-MT benefit from mutual learning which help the model to achieve better performance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

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

Song

Zhang

2023

J. Chem. Theory Comput.

View full text Add to dashboard Cite

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.

show abstract

“…Results from previous SAMPL log P challenges have reported that when a large experimental data set is available, data-driven methods like ML approaches can prove to be more accurate and time efficient than CSM based methods. [5][6][7][8][9] However, the unavailability of sufficient and comprehensive experimental data for log P values poses a major problem. Insufficient data for these methods can make them less reliable and reduce their generalizability.…”

Section: Introductionmentioning

confidence: 99%

Blind prediction of toluene/water partition coefficients using COSMO-RS: results from the SAMPL9 challenge

Nevolianis,

Ahmed,

Hellweg

et al. 2023

Phys. Chem. Chem. Phys.

View full text Add to dashboard Cite

show abstract

COSMO-RS predictions of logP in the SAMPL7 blind challenge

Cited by 14 publications

References 23 publications

Density Functional Theory with Implicit Solvents for Accurate Estimation of Aqueous and Organic Solvation Free Energies of Lignin Fragments

Density Functional Theory with Implicit Solvents for Accurate Estimation of Aqueous and Organic Solvation Free Energies of Lignin Fragments

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

Blind prediction of toluene/water partition coefficients using COSMO-RS: results from the SAMPL9 challenge

Contact Info

Product

Resources

About