3D matters! 3D-RISM and 3D convolutional neural network for accurate bioaccumulation prediction

Sergey Sosnin, Maksim Misin, David S. Palmer, Maxim V. Fedorov

Research output: Contribution to journalArticle

Abstract

In this work, we present a new method for predicting complex physicalchemical properties of organic molecules. The approach utilizes 3D convolutional neural network (ActivNet4) that uses solvent spatial distributions around solutes as input. These spatial distributions are obtained by a molecular theory called threedimensional reference interaction site model (3D-RISM). We have shown that the method allows one to achieve a good accuracy of prediction of bioconcentration factor (BCF) which is difficult to predict by direct application of methods of molecular theory or simulations. Our research demonstrates that combination of molecular theories with modern machine learning approaches can be effectively used for predicting properties that are otherwise inaccessible to purely theory-based models.
LanguageEnglish
Article number32LT03
Number of pages7
JournalJournal of Physics: Condensed Matter
Volume30
DOIs
StatePublished - 19 Jul 2018

Fingerprint

molecular theory
Bioaccumulation
Spatial distribution
Neural networks
spatial distribution
predictions
Learning systems
machine learning
Molecules
solutes
molecules
simulation
interactions

Keywords

  • deep learning (DL)
  • convolution neural network (CNN)
  • 3DRISM
  • machine learning
  • bioaccumulation
  • molecular solvation
  • computational chemistry
  • molecular simulation

Cite this

@article{c623c196732b4cc9a27644566f37c555,
title = "3D matters! 3D-RISM and 3D convolutional neural network for accurate bioaccumulation prediction",
abstract = "In this work, we present a new method for predicting complex physicalchemical properties of organic molecules. The approach utilizes 3D convolutional neural network (ActivNet4) that uses solvent spatial distributions around solutes as input. These spatial distributions are obtained by a molecular theory called threedimensional reference interaction site model (3D-RISM). We have shown that the method allows one to achieve a good accuracy of prediction of bioconcentration factor (BCF) which is difficult to predict by direct application of methods of molecular theory or simulations. Our research demonstrates that combination of molecular theories with modern machine learning approaches can be effectively used for predicting properties that are otherwise inaccessible to purely theory-based models.",
keywords = "deep learning (DL), convolution neural network (CNN), 3DRISM, machine learning, bioaccumulation, molecular solvation, computational chemistry, molecular simulation",
author = "Sergey Sosnin and Maksim Misin and Palmer, {David S.} and Fedorov, {Maxim V.}",
year = "2018",
month = "7",
day = "19",
doi = "10.1088/1361-648X/aad076",
language = "English",
volume = "30",
journal = "Journal of Physics: Condensed Matter",
issn = "0953-8984",

}

3D matters! 3D-RISM and 3D convolutional neural network for accurate bioaccumulation prediction. / Sosnin, Sergey; Misin, Maksim; Palmer, David S.; Fedorov, Maxim V.

In: Journal of Physics: Condensed Matter, Vol. 30, 32LT03 , 19.07.2018.

Research output: Contribution to journalArticle

TY - JOUR

T1 - 3D matters! 3D-RISM and 3D convolutional neural network for accurate bioaccumulation prediction

AU - Sosnin,Sergey

AU - Misin,Maksim

AU - Palmer,David S.

AU - Fedorov,Maxim V.

PY - 2018/7/19

Y1 - 2018/7/19

N2 - In this work, we present a new method for predicting complex physicalchemical properties of organic molecules. The approach utilizes 3D convolutional neural network (ActivNet4) that uses solvent spatial distributions around solutes as input. These spatial distributions are obtained by a molecular theory called threedimensional reference interaction site model (3D-RISM). We have shown that the method allows one to achieve a good accuracy of prediction of bioconcentration factor (BCF) which is difficult to predict by direct application of methods of molecular theory or simulations. Our research demonstrates that combination of molecular theories with modern machine learning approaches can be effectively used for predicting properties that are otherwise inaccessible to purely theory-based models.

AB - In this work, we present a new method for predicting complex physicalchemical properties of organic molecules. The approach utilizes 3D convolutional neural network (ActivNet4) that uses solvent spatial distributions around solutes as input. These spatial distributions are obtained by a molecular theory called threedimensional reference interaction site model (3D-RISM). We have shown that the method allows one to achieve a good accuracy of prediction of bioconcentration factor (BCF) which is difficult to predict by direct application of methods of molecular theory or simulations. Our research demonstrates that combination of molecular theories with modern machine learning approaches can be effectively used for predicting properties that are otherwise inaccessible to purely theory-based models.

KW - deep learning (DL)

KW - convolution neural network (CNN)

KW - 3DRISM

KW - machine learning

KW - bioaccumulation

KW - molecular solvation

KW - computational chemistry

KW - molecular simulation

UR - http://iopscience.iop.org/journal/0953-8984

U2 - 10.1088/1361-648X/aad076

DO - 10.1088/1361-648X/aad076

M3 - Article

VL - 30

JO - Journal of Physics: Condensed Matter

T2 - Journal of Physics: Condensed Matter

JF - Journal of Physics: Condensed Matter

SN - 0953-8984

M1 - 32LT03

ER -