Machine learning prediction of 3CL  SARS-CoV-2 docking scores

Bučinský, Lukáš; Bortňák, Dušan; Gall, Marián; Matúška, Ján; Milata, Viktor; Pitoňák, Michal; Štekláč, Marek; Végh, Dániel; Zajaček, Dávid

doi:10.1016/j.compbiolchem.2022.107656

Cited by 11 publications

(12 citation statements)

References 102 publications

(129 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The absolute error for this compound is 8.683, 5.084 and 14.397 kcal/mol for TensorFlow, XGBoost and SchNetPack, respectively. The prediction of docking scores of compounds with >120 atoms was found problematic in the original work of Bucinsky et al [ 26 ], which is confirmed herein for the case of the tannic acid. The agreement between the AutoDock calculated (expected) and ML predicted docking scores improves after leaving this compound out of the linear correlation, see Table 2 .…”

Section: Resultssupporting

confidence: 60%

“…The protease was stripped of water molecules, retaining only a single water molecule near His41 and Asp187 residues facilitating charge stabilization interactions on nearby residues [ 32 ]. The pdbqt file format of the 6WQF protein structure for the W set docking study has been prepared with AutoDockTools-1.5.7 as given by Bucinsky et al [ 26 ] (keeping only polar hydrogens and using Gasteiger charges). Semi-flexible molecular docking calculations were performed using Autodock4.2.6 [ 34 , 35 ] with compounds having assigned Gasteiger charges.…”

Section: Methods and Computational Detailsmentioning

confidence: 99%

“…Machine learning predictions of the docking scores were carried out using TensorFlow [ 42 ], XGBoost [ 43 ], and SchNetPack [ 44 ] utilizing the S training set for the ML process itself according to Bucinsky et al [ 26 ]. All W set docking scores, and their ML predictions are accounted for in the Supplementary Materials.…”

Section: Methods and Computational Detailsmentioning

confidence: 99%

“…Molecular docking became the standard approach to screen for possibly active substances for a further evaluation by means of more demanding molecular dynamics simulations or further in vitro tests in a targeted drug development. Molecular docking studies of large data sets of compounds were reported [ 13 , [24] , [25] , [26] , [27] ]. Nevertheless, even the molecular docking step is demanding when a quick tool for shrinking a database of thousands of compounds to a reasonable time is desirable [ 24 , 25 , 28 ].…”

Section: Introductionmentioning

confidence: 99%

“…Docking scores of five best compounds are further validated by means of molecular dynamics. Subsequently, the performance of ML docking scores prediction is evaluated according to Bucinsky et al [ 26 ]. The accuracy of the ML predicted docking scores of W set compounds is presented and the time scales (docking vs. ML) are presented.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Molecular docking and machine learning affinity prediction of compounds identified upon softwood bark extraction to the main protease of the SARS-CoV-2 virus

Jablonský

Štekláč

Majová

et al. 2022

Biophysical Chemistry

Self Cite

View full text Add to dashboard Cite

Section: Resultssupporting

confidence: 60%

Section: Methods and Computational Detailsmentioning

confidence: 99%

Section: Methods and Computational Detailsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Molecular docking and machine learning affinity prediction of compounds identified upon softwood bark extraction to the main protease of the SARS-CoV-2 virus

Jablonský

Štekláč

Majová

et al. 2022

Biophysical Chemistry

Self Cite

View full text Add to dashboard Cite

Advances and critical assessment of machine learning techniques for prediction of docking scores

Bučinský

Gall

Matúška

et al. 2023

Int J of Quantum Chemistry

Self Cite

View full text Add to dashboard Cite

Here we present three distinct machine learning (ML) approaches (TensorFlow, XGBoost, and SchNetPack) for docking score prediction. AutoDock Vina is used to evaluate the inhibitory potential of ZINC15 in‐vivo and in‐vitro‐only sets towards the SARS‐CoV‐2 main protease. The in‐vivo set (59 884 compounds) is used for ML training (max. 80%), validation (5%), and testing (15%). The in‐vitro‐only set (174 014 compounds) is used for the evaluation of prediction capability of the trained ML models. Contributions to the prediction error are analyzed with respect to compounds' charge, number of atoms, and expected inhibitory potential (docking score). Methods for the prediction error estimation of new compounds are considered, yet critically rejected. The ML input weighted with respect to the desired property (i.e., low docking score) in the machine learning models shows to be a promising option to improve the ML performance. Proposed models provide significant reduction in number of intriguing compounds that need to be investigated.

show abstract

Improving drug discovery with a hybrid deep generative model using reinforcement learning trained on a Bayesian docking approximation

Xiong,

Wang,

Wang

et al. 2023

J Comput Aided Mol Des

View full text Add to dashboard Cite

Machine learning prediction of 3CL SARS-CoV-2 docking scores

Cited by 11 publications

References 102 publications

Molecular docking and machine learning affinity prediction of compounds identified upon softwood bark extraction to the main protease of the SARS-CoV-2 virus

Molecular docking and machine learning affinity prediction of compounds identified upon softwood bark extraction to the main protease of the SARS-CoV-2 virus

Advances and critical assessment of machine learning techniques for prediction of docking scores

Improving drug discovery with a hybrid deep generative model using reinforcement learning trained on a Bayesian docking approximation

Contact Info

Product

Resources

About