Cargando…

Predicting chemical ecotoxicity by learning latent space chemical representations

In silico prediction of chemical ecotoxicity (HC(50)) represents an important complement to improve in vivo and in vitro toxicological assessment of manufactured chemicals. Recent application of machine learning models to predict chemical HC(50) yields variable prediction performance that depends on...

Descripción completa

Detalles Bibliográficos
Autores principales: Gao, Feng, Zhang, Wei, Baccarelli, Andrea A., Shen, Yike
Formato: Online Artículo Texto
Lenguaje:English
Publicado: 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9044254/
https://www.ncbi.nlm.nih.gov/pubmed/35395577
http://dx.doi.org/10.1016/j.envint.2022.107224
_version_ 1784695065719865344
author Gao, Feng
Zhang, Wei
Baccarelli, Andrea A.
Shen, Yike
author_facet Gao, Feng
Zhang, Wei
Baccarelli, Andrea A.
Shen, Yike
author_sort Gao, Feng
collection PubMed
description In silico prediction of chemical ecotoxicity (HC(50)) represents an important complement to improve in vivo and in vitro toxicological assessment of manufactured chemicals. Recent application of machine learning models to predict chemical HC(50) yields variable prediction performance that depends on effectively learning chemical representations from high-dimension data. To improve HC(50) prediction performance, we developed an autoencoder model by learning latent space chemical embeddings. This novel approach achieved state-of-the-art prediction performance of HC(50) with R(2) of 0.668 ± 0.003 and mean absolute error (MAE) of 0.572 ± 0.001, and outperformed other dimension reduction methods including principal component analysis (PCA) (R(2) = 0.601 ± 0.031 and MAE = 0.629 ± 0.005), kernel PCA (R(2) = 0.631 ± 0.008 and MAE = 0.625 ± 0.006), and uniform manifold approximation and projection dimensionality reduction (R(2) = 0.400 ± 0.008 and MAE = 0.801 ± 0.002). A simple linear layer with chemical embeddings learned from the autoencoder model performed better than random forest (R(2) = 0.663 ± 0.007 and MAE = 0.591 ± 0.008), fully connected neural network (R(2) = 0.614 ± 0.016 and MAE = 0.610 ± 0.008), least absolute shrinkage and selection operator (R(2) = 0.617 ± 0.037 and MAE = 0.619 ± 0.007), and ridge regression (R(2) = 0.638 ± 0.007 and MAE = 0.613 ± 0.005) using unlearned raw input features. Our results highlighted the usefulness of learning latent chemical representations, and our autoencoder model provides an alternative approach for robust HC(50) prediction.
format Online
Article
Text
id pubmed-9044254
institution National Center for Biotechnology Information
language English
publishDate 2022
record_format MEDLINE/PubMed
spelling pubmed-90442542022-05-01 Predicting chemical ecotoxicity by learning latent space chemical representations Gao, Feng Zhang, Wei Baccarelli, Andrea A. Shen, Yike Environ Int Article In silico prediction of chemical ecotoxicity (HC(50)) represents an important complement to improve in vivo and in vitro toxicological assessment of manufactured chemicals. Recent application of machine learning models to predict chemical HC(50) yields variable prediction performance that depends on effectively learning chemical representations from high-dimension data. To improve HC(50) prediction performance, we developed an autoencoder model by learning latent space chemical embeddings. This novel approach achieved state-of-the-art prediction performance of HC(50) with R(2) of 0.668 ± 0.003 and mean absolute error (MAE) of 0.572 ± 0.001, and outperformed other dimension reduction methods including principal component analysis (PCA) (R(2) = 0.601 ± 0.031 and MAE = 0.629 ± 0.005), kernel PCA (R(2) = 0.631 ± 0.008 and MAE = 0.625 ± 0.006), and uniform manifold approximation and projection dimensionality reduction (R(2) = 0.400 ± 0.008 and MAE = 0.801 ± 0.002). A simple linear layer with chemical embeddings learned from the autoencoder model performed better than random forest (R(2) = 0.663 ± 0.007 and MAE = 0.591 ± 0.008), fully connected neural network (R(2) = 0.614 ± 0.016 and MAE = 0.610 ± 0.008), least absolute shrinkage and selection operator (R(2) = 0.617 ± 0.037 and MAE = 0.619 ± 0.007), and ridge regression (R(2) = 0.638 ± 0.007 and MAE = 0.613 ± 0.005) using unlearned raw input features. Our results highlighted the usefulness of learning latent chemical representations, and our autoencoder model provides an alternative approach for robust HC(50) prediction. 2022-05 2022-04-01 /pmc/articles/PMC9044254/ /pubmed/35395577 http://dx.doi.org/10.1016/j.envint.2022.107224 Text en https://creativecommons.org/licenses/by-nc-nd/4.0/This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/ (https://creativecommons.org/licenses/by-nc-nd/4.0/) ).
spellingShingle Article
Gao, Feng
Zhang, Wei
Baccarelli, Andrea A.
Shen, Yike
Predicting chemical ecotoxicity by learning latent space chemical representations
title Predicting chemical ecotoxicity by learning latent space chemical representations
title_full Predicting chemical ecotoxicity by learning latent space chemical representations
title_fullStr Predicting chemical ecotoxicity by learning latent space chemical representations
title_full_unstemmed Predicting chemical ecotoxicity by learning latent space chemical representations
title_short Predicting chemical ecotoxicity by learning latent space chemical representations
title_sort predicting chemical ecotoxicity by learning latent space chemical representations
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9044254/
https://www.ncbi.nlm.nih.gov/pubmed/35395577
http://dx.doi.org/10.1016/j.envint.2022.107224
work_keys_str_mv AT gaofeng predictingchemicalecotoxicitybylearninglatentspacechemicalrepresentations
AT zhangwei predictingchemicalecotoxicitybylearninglatentspacechemicalrepresentations
AT baccarelliandreaa predictingchemicalecotoxicitybylearninglatentspacechemicalrepresentations
AT shenyike predictingchemicalecotoxicitybylearninglatentspacechemicalrepresentations