Cargando…
QSAR with experimental and predictive distributions: an information theoretic approach for assessing model quality
We propose that quantitative structure–activity relationship (QSAR) predictions should be explicitly represented as predictive (probability) distributions. If both predictions and experimental measurements are treated as probability distributions, the quality of a set of predictive distributions out...
Autores principales: | , , , , |
---|---|
Formato: | Online Artículo Texto |
Lenguaje: | English |
Publicado: |
Springer Netherlands
2013
|
Materias: | |
Acceso en línea: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3639359/ https://www.ncbi.nlm.nih.gov/pubmed/23504478 http://dx.doi.org/10.1007/s10822-013-9639-5 |
_version_ | 1782475945110667264 |
---|---|
author | Wood, David J. Carlsson, Lars Eklund, Martin Norinder, Ulf Stålring, Jonna |
author_facet | Wood, David J. Carlsson, Lars Eklund, Martin Norinder, Ulf Stålring, Jonna |
author_sort | Wood, David J. |
collection | PubMed |
description | We propose that quantitative structure–activity relationship (QSAR) predictions should be explicitly represented as predictive (probability) distributions. If both predictions and experimental measurements are treated as probability distributions, the quality of a set of predictive distributions output by a model can be assessed with Kullback–Leibler (KL) divergence: a widely used information theoretic measure of the distance between two probability distributions. We have assessed a range of different machine learning algorithms and error estimation methods for producing predictive distributions with an analysis against three of AstraZeneca’s global DMPK datasets. Using the KL-divergence framework, we have identified a few combinations of algorithms that produce accurate and valid compound-specific predictive distributions. These methods use reliability indices to assign predictive distributions to the predictions output by QSAR models so that reliable predictions have tight distributions and vice versa. Finally we show how valid predictive distributions can be used to estimate the probability that a test compound has properties that hit single- or multi- objective target profiles. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1007/s10822-013-9639-5) contains supplementary material, which is available to authorized users. |
format | Online Article Text |
id | pubmed-3639359 |
institution | National Center for Biotechnology Information |
language | English |
publishDate | 2013 |
publisher | Springer Netherlands |
record_format | MEDLINE/PubMed |
spelling | pubmed-36393592013-04-30 QSAR with experimental and predictive distributions: an information theoretic approach for assessing model quality Wood, David J. Carlsson, Lars Eklund, Martin Norinder, Ulf Stålring, Jonna J Comput Aided Mol Des Article We propose that quantitative structure–activity relationship (QSAR) predictions should be explicitly represented as predictive (probability) distributions. If both predictions and experimental measurements are treated as probability distributions, the quality of a set of predictive distributions output by a model can be assessed with Kullback–Leibler (KL) divergence: a widely used information theoretic measure of the distance between two probability distributions. We have assessed a range of different machine learning algorithms and error estimation methods for producing predictive distributions with an analysis against three of AstraZeneca’s global DMPK datasets. Using the KL-divergence framework, we have identified a few combinations of algorithms that produce accurate and valid compound-specific predictive distributions. These methods use reliability indices to assign predictive distributions to the predictions output by QSAR models so that reliable predictions have tight distributions and vice versa. Finally we show how valid predictive distributions can be used to estimate the probability that a test compound has properties that hit single- or multi- objective target profiles. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1007/s10822-013-9639-5) contains supplementary material, which is available to authorized users. Springer Netherlands 2013-03-16 2013 /pmc/articles/PMC3639359/ /pubmed/23504478 http://dx.doi.org/10.1007/s10822-013-9639-5 Text en © The Author(s) 2013 https://creativecommons.org/licenses/by/2.0/ Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited. |
spellingShingle | Article Wood, David J. Carlsson, Lars Eklund, Martin Norinder, Ulf Stålring, Jonna QSAR with experimental and predictive distributions: an information theoretic approach for assessing model quality |
title | QSAR with experimental and predictive distributions: an information theoretic approach for assessing model quality |
title_full | QSAR with experimental and predictive distributions: an information theoretic approach for assessing model quality |
title_fullStr | QSAR with experimental and predictive distributions: an information theoretic approach for assessing model quality |
title_full_unstemmed | QSAR with experimental and predictive distributions: an information theoretic approach for assessing model quality |
title_short | QSAR with experimental and predictive distributions: an information theoretic approach for assessing model quality |
title_sort | qsar with experimental and predictive distributions: an information theoretic approach for assessing model quality |
topic | Article |
url | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3639359/ https://www.ncbi.nlm.nih.gov/pubmed/23504478 http://dx.doi.org/10.1007/s10822-013-9639-5 |
work_keys_str_mv | AT wooddavidj qsarwithexperimentalandpredictivedistributionsaninformationtheoreticapproachforassessingmodelquality AT carlssonlars qsarwithexperimentalandpredictivedistributionsaninformationtheoreticapproachforassessingmodelquality AT eklundmartin qsarwithexperimentalandpredictivedistributionsaninformationtheoreticapproachforassessingmodelquality AT norinderulf qsarwithexperimentalandpredictivedistributionsaninformationtheoreticapproachforassessingmodelquality AT stalringjonna qsarwithexperimentalandpredictivedistributionsaninformationtheoreticapproachforassessingmodelquality |