ABSTRACT
Introduction
The application of Artificial Intelligence (AI) to predictive toxicology is rapidly increasing, particularly aiming to develop non-testing methods that effectively address ethical concerns and reduce economic costs. In this context, Developmental Toxicity (Dev Tox) stands as a key human health endpoint, especially significant for safeguarding maternal and child well-being.
Areas covered
This review outlines the existing methods employed in Dev Tox predictions and underscores the benefits of utilizing New Approach Methodologies (NAMs), specifically focusing on eXplainable Artificial Intelligence (XAI), which proves highly efficient in constructing reliable and transparent models aligned with recommendations from international regulatory bodies.
Expert opinion
The limited availability of high-quality data and the absence of dependable Dev Tox methodologies render XAI an appealing avenue for systematically developing interpretable and transparent models, which hold immense potential for both scientific evaluations and regulatory decision-making.
Article highlights
Developmental toxicity is a critical human health endpoint, essential for safeguarding maternal and child well-being.
Explainable artificial intelligence facilitates the creation of transparent and interpretable predictive models.
Non-testing methods alleviate both ethical and economic concerns.
Collecting a greater volume of high-quality Dev Tox data is imperative to prevent misleading results.
Reliable and transparent predictions qualify models for regulatory use.
Abbreviations
ACToR | = | Aggregated Computational Toxicology Resource |
AD | = | Applicability Domain |
ADMET | = | absorption, distribution, metabolism, excretion, and toxicity |
ADR | = | Adverse Drug Reaction |
AI | = | Artificial intelligence |
ANNs | = | Artificial Neural Networks |
AUC | = | Area Under Curve |
BPNNs | = | Back Propagation Neural Networks |
BRANNs | = | Bayesian-regularized Neural Networks |
CA | = | Classification Accuracy |
CADDementia | = | Computer-Aided Diagnosis of Dementia |
CAESAR | = | Computer Assisted Evaluation of industrial chemical Substances According to Regulations |
CART | = | Classification and Regression Tree |
CDC | = | Centers for Disease Control |
CEPA | = | Canadian Environmental Protection Act |
ChEMBL | = | Chemical European Molecular Biology Laboratory |
CLP | = | Classification, Labelling and Packaging |
CV | = | Cross-Validation |
DART | = | Development and Reproductive Toxicology |
DB-ALM | = | Database on Alternative Methods to Animal Experimentation |
Dev Tox | = | Developmental Toxicity |
DL | = | Deep Learning |
DNNs | = | Deep Neural Networks |
DSSTox | = | Distributed Structure-Searchable Toxicity |
DT | = | Decision Tree |
ECHA | = | European Chemical Agency |
ECVAM | = | European Committee for the Validation of Alternative Methods |
EMA | = | European Medicinal Chemistry |
EPA | = | Environmental Protection Agency |
EST | = | Embryonic Stem cell Test |
FAIR | = | Findable, Accessible, Interoperable and Reusable |
FDA | = | Food and Drug Administration |
FP | = | FingerPrint |
GLP | = | Good Laboratory Practice |
GSH | = | Globally Harmonized System of Classification and Labelling of Chemicals |
HTS | = | High-Throughput Screening |
ICH | = | International Council for Harmonization |
k-NN | = | k-Nearest Neighbors |
LD50 | = | Lethal Dose 50 |
LOAEL | = | Lowest Observed Adverse Effect Level |
LR | = | Linear Regression |
MACCS | = | Molecular ACCess System |
MACCSFP | = | Molecular ACCess Systems FingerPrint |
MCC | = | Matthews Correlation Coefficient |
MCM | = | Major Congenital Malformations |
ML | = | Machine Learning |
MLR | = | Multiple Linear Regression |
MM | = | micromass |
MOE | = | Molecular Operating Environment |
NAM | = | New Approach Methodologies |
NBC | = | Naïve Bayes Classifier |
NCCT | = | National Center for Computational Toxicology |
NCE | = | New Chemical Entity |
NICETAM | = | NTP Interagency Center for the Evaluation of Alternative Toxicological Methods |
NIH | = | National Institute of Health |
NLM | = | National Library of Medicine |
NLP | = | Natural Language Processing |
NOAEL | = | No Observed Adverse Effect Level |
NTP | = | National Toxicology Program |
OECD | = | Organization for Economic Co-operation and Development |
OPS | = | Optimum Prediction Space |
P&G | = | Procter and Gamble |
PAH | = | Polycyclic Aromatic Hydrocarbons |
PCA | = | Principal Component Analysis |
PLS | = | Partial Least Square |
qed | = | quantification of estimation of drug-likeness |
QSAR | = | Quantitative Structure-Activity Relationship |
RDT | = | Repeated-Dose Toxicity |
REACH | = | Registration, Evaluation, Authorization and restriction of Chemicals |
Repro Tox | = | Reproductive Toxicology Center System |
RF | = | Random Forest |
SAR | = | Structure-Activity Relationship |
SEAZIT | = | Systematic Evaluation of the Application of Zebrafish in Toxicology |
SHAP | = | Shapley Additive exPlanation |
SMARTS | = | SMILES arbitrary target specification |
SMILES | = | Simplified Molecular Input Line Entry System |
SVM | = | Support Vector Machine |
T.E.S.T. | = | Toxicity Estimation Software Tool |
TERIS | = | Teratogen Information System |
TG | = | Test Guidelines |
TIRESIA | = | Toxicology Intelligence and Regulatory Evaluations for Scientific and Industry Applications |
ToxNET | = | Toxicology Data Network |
ToxRefDB | = | Toxicity Reference Database |
US EPA | = | U.S. Environmental Protection Agency |
WEC | = | Embryo Culture test |
WHO | = | World Health Organization |
XAI | = | eXplainable Artificial Intelligence |
XGB | = | eXtreme Gradient Boosting |
Declaration of interests
The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest or conflict with the subject matter or materials discussed in this manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, and royalties.
Reviewer disclosures
Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.
Acknowledgments
The authors are thankful to the following: Agenzia Regionale Strategica per la Salute e il Sociale (A.Re.S.S. PUGLIA); Programma Operativo Nazionale Ricerca e Innovazione 2014-2020 (CCI 2014IT16M2OP005), Fondo Sociale Europeo, Azione I.1 “Dottorati Innovativi con caratterizzazione Industriale” the Comune di Bovino and Piano Stralcio “Ricerca e Innovazione” 2015-2017, Comune di Bovino (Foggia, Italy) and Dott. Renato Lombardi, Director of “Struttura Complessa di Farmacia, IRCCS, Casa Sollievo della Sofferenza”, (Foggia, Italy) (Code: DOT19C9KX4); Horizon Europe Seeds “L’intelligenza artificiale a tutela della salute in età pediatrica. Implementazione di una piattaforma digitale per il design di farmaci pediatrici sicuri”, Università degli Studi di Bari (Bari, Italy) (CUP: H99J21017390006).