Machine Learning and Deep Learning in Computational Toxicology

654 Pages·2023·19.482 MB·English

by  Huixiao Hong

Computational Methods in Engineering & the Sciences
Huixiao Hong   Editor
Machine
Learning and
Deep Learning in
Computational
Toxicology
Computational Methods in Engineering &
the Sciences
Series Editor
Klaus-Jürgen Bathe, Department of Mechanical Engineering, Massachusetts
Institute of Technology, Cambridge, MA, USA
This Series publishes books on all aspects of computational methods used in
engineering and the sciences. With emphasis on simulation through mathematical
modelling, the Series accepts high quality content books across different domains of
engineering, materials, and other applied sciences. The Series publishes mono-
graphs, contributed volumes, professional books, and handbooks, spanning across
cutting edge research as well as basics of professional practice. The topics of
interest include the development and applications of computational simulations in
the broad fields of Solid & Structural Mechanics, Fluid Dynamics, Heat Transfer,
Electromagnetics, Multiphysics, Optimization, Stochastics with simulations in and
for Structural Health Monitoring, Energy Systems, Aerospace Systems, Machines
and Turbines. Climate Prediction, Effects of Earthquakes, Geotechnical Systems,
Chemical and Biomolecular Systems, Molecular Biology, Nano and Microfluidics,
Materials Science, Nanotechnology, Manufacturing and 3D printing, Artificial
Intelligence, Internet-of-Things.
Huixiao Hong
Editor
Machine Learning and Deep
Learning in Computational
Toxicology
Editor
Huixiao Hong
Division of Bioinformatics and Biostatistics
National Center for Toxicological Research
U.S. Food and Drug Administration
Jefferson, AR, USA
ISSN 2662-4869 ISSN 2662-4877 (electronic)
Computational Methods in Engineering & the Sciences
ISBN 978-3-031-20729-7 ISBN 978-3-031-20730-3 (eBook)
https://doi.org/10.1007/978-3-031-20730-3
This is a U.S. government work and not under copyright protection in the U.S.; foreign copyright
protection may apply 2023
All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilms or in any other physical way, and transmission or information storage and
retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known
or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Owing to the recent explosive growth of the chemical industry, compounded by
improved methods to rapidly synthesize compounds with more diverse chemical
structures, the need to efficiently assess the risk to public health of chemicals and
evaluate the safety profile of chemical containing products is imperative. Safety
evaluation and risk assessment are needed to protect not only the human species,
but also other living organisms and our ecosystem. In addition to merely needing to
assess chemicals, the field must also assess byproducts of chemicals and interactions
with other chemicals that can occur in living organisms or the environment.
Until recently, the gold standard for toxicology hazard assessment was animal
models. However, animal models have numerous fundamental limitations around
their utility, including, but not limited to, high cost, low throughput, low accuracy,
and ethical concerns. Thus, the scientific community as a whole has declared the
importance of the three “Rs”: Replace, Reduce, and Refine animal use. To meet
any of these three R goals, alternative methods are required to assess the safety and
hazard of chemicals.
Fortunately, the toxicology community is beginning to fully realize the issues
associated with animal models concomitant with incredible advancement in the
field of computer science. Thus, computational toxicologists are collaborating with
computer scientists to spark the dawn of the computational toxicology era. This field
has the goal of predicting and understanding the hazards of chemicals in a cost-and
time-efficient manner with very low risk of ethical issues.
As new computational methods are developed, they can be applied to toxicology
challenges, which will allow for continual iteration and improvement. Currently,
artificial intelligence, mainly machine learning and deep learning, is in a Cambrian
era with new advanced algorithms being developed and deployed across a number of
fields. This rapid advance has quickly transformed industry, jobs, and even society.
While machine learning and deep learning were developed in the context of other
fields, they have all been successfully applied to computational toxicology.
v
vi Preface
This textbook will survey the landscape of the deployment of various advanced
algorithms to solve important questions around the hazard and toxicity of chemi-
cals. As many of these algorithms come from other fields, this work will also cover
important best practices to be aware while deploying such algorithms to solve crit-
ical challenges. To give well-deserved attention to the many facets of computational
toxicology, this book is divided into four parts. First, the text will detail the machine
learning and deep learning algorithms most relevant to the field of computational
toxicology and why they are relevant. Next, this will be expanded by describing
tools and approaches to enable efficient application of advanced algorithms to toxi-
cology. As examples are important to illustrate relevance, this textbook will then
focus on the application of machine learning and deep learning to chemical toxicity
prediction. This textbook will also delve into the future of nascent approaches and
new progress in the area.
This textbook was authored with a wide audience in mind. For the classical toxicol-
ogist looking to dip a toe into computational approaches, computational concepts are
thoroughly introduced and explained through clarifying examples. For the computer
scientist looking to apply machine learning and deep learning to other fields, this
textbook showcases numerous examples of how others have successfully applied
advanced computation to solve toxicology problems. Lastly, for students and other
trainees eager to further their career, this body of work will survey a variety of
computational toxicology topics and could provide a spark of inspiration for a career
direction.
While the chapters following do review a variety of computational toxicology
topics, it does not contain computer code samples, chapter quizzes, or practice
exams. For instructional purposes, developing quizzes and exams are left to a
course instructor. The authors would humbly appreciate comments, feedback, and
corrections from readers so future work can be improved.
The impressive quantity, speed, and diversity of data relevant to computational
toxicology have substantially grown in recent years. These increasingly rich datasets
have fueled the growth of machine learning and deep learning approaches, which are
hungry for large and high-quality datasets. As the fields of toxicology and computer
science continue to intersect and synchronize in new ways, the world will benefit
from improved prediction and understanding of the toxicity and hazard potential of
chemicals.
Preface vii
This preface reflects the views of the authors and does not necessarily reflect those
of the U.S. Food and Drug Administration.
Rebecca Kusko, Ph.D.
Head of Business Development and
Corporate Affairs
Immuneering Corporation
Cambridge, MA, USA
Huixiao Hong, Ph.D.
SBRBPAS Expert
Chief
Bioinformatics Branch
National Center for Toxicological Research
U.S. Food and Drug Administration
Jefferson, AR, USA
Contents
1 Machine Learning and Deep Learning Promote
Computational Toxicology for Risk Assessment of Chemicals ...... 1
Rebecca Kusko and Huixiao Hong
Part I Machine Learning and Deep Learning Methods for
Computational Toxicology
2 Assessment of the Xenobiotics Toxicity Taking into Account
Their Metabolism ............................................. 21
Dmitry Filimonov, Alexander Dmitriev, Anastassia Rudik,
and Vladimir Poroikov
3 Emerging Machine Learning Techniques in Predicting
Adverse Drug Reactions ....................................... 53
Yi Zhong, Shanshan Wang, Gaozheng Li, Ji Yang, Zuquan Weng,
and Heng Luo
4 Drug Effect Deep Learner Based on Graphical Convolutional
Network ...................................................... 83
Yunyi Wu, Shenghui Guan, and Guanyu Wang
5 AOP-Based Machine Learning for Toxicity Prediction ............ 141
Wei Shi, Rong Zhang, and Haoyue Tan
6 Graph Kernel Learning for Predictive Toxicity Models ........... 159
Youjun Xu, Chia-Han Chou, Ningsheng Han, Jianfeng Pei,
and Luhua Lai
7 Optimize and Strengthen Machine Learning Models
Based on in Vitro Assays with Mechanistic Knowledge
and Real-World Data .......................................... 183
Thilini V. Mahanama, Arpan Biswas, and Dong Wang
ix
x Contents
8 Multitask Learning for Quantitative Structure–Activity
Relationships: A Tutorial ...................................... 199
Cecile Valsecchi, Francesca Grisoni, Viviana Consonni,
Davide Ballabio, and Roberto Todeschini
Part II Tools and Approaches Facilitating Machine Learning
and Deep Learning Methods in Computational Toxicology
9 Isalos Predictive Analytics Platform: Cheminformatics,
Nanoinformatics, and Data Mining Applications ................. 223
Dimitra-Danai Varsou, Andreas Tsoumanis,
Anastasios G. Papadiamantis, Georgia Melagraki,
and Antreas Afantitis
10 ED Profiler: Machine Learning Tool for Screening Potential
Endocrine-Disrupting Chemicals ............................... 243
Xianhai Yang, Huihui Liu, Rebecca Kusko, and Huixiao Hong
11 Quantitative Target-specific Toxicity Prediction Modeling
(QTTPM): Coupling Machine Learning with Dynamic
Protein–Ligand Interaction Descriptors (DyPLIDs) to Predict
Androgen Receptor-mediated Toxicity .......................... 263
Sundar Thangapandian, Gabriel Idakwo, Joseph Luttrell,
Huixiao Hong, Chaoyang Zhang, and Ping Gong
12 Mold2 Descriptors Facilitate Development of Machine
Learning and Deep Learning Models for Predicting Toxicity
of Chemicals .................................................. 297
Huixiao Hong, Jie Liu, Weigong Ge, Sugunadevi Sakkiah,
Wenjing Guo, Gokhan Yavas, Chaoyang Zhang, Ping Gong,
Weida Tong, and Tucker A. Patterson
13 Applicability Domain Characterization for Machine Learning
QSAR Models ................................................ 323
Zhongyu Wang and Jingwen Chen
14 Controlling for Confounding in Complex Survey Machine
Learning Models to Assess Drug Safety and Risk ................. 355
Paul Rogers
15 Multivariate Curve Resolution for Analysis of Heterogeneous
System in Toxicogenomics ...................................... 375
Yuan Liu, Jinzhu Lin, Menglong Li, and Zhining Wen