Molecular similarity in chemical informatics and predictive toxicity modeling: from quantitative read-across (q-RA) to quantitative read-across structure–activity relationship (q-RASAR) with the application of machine learning

Critical Reviews in Toxicology, Journal Year: 2024, Volume and Issue: 54(9), P. 659 - 684

Published: Sept. 3, 2024

This article aims to provide a comprehensive critical, yet readable, review of general interest the chemistry community on molecular similarity as applied chemical informatics and predictive modeling with special focus read-across (RA) structure-activity relationships (RASAR). Molecular similarity-based computational tools, such quantitative (QSARs) RA, are routinely used fill data gaps for wide range properties including toxicity endpoints regulatory purposes. will explore background RA starting from how structural information has been through other contexts physicochemical, absorption, distribution, metabolism, elimination (ADME) properties, biological aspects being characterized. More recent developments RA's integration QSAR have resulted in emergence novel models ToxRead, generalized (GenRA), RASAR (q-RASAR). Conventional techniques excluded this except where necessary context.

Language: Английский

---

Integrating Machine Learning and Nano-QSAR Models to Predict the Oxidative Stress Potential Caused by Single and Mixed Carbon Nanomaterials in Algal Cells

Environmental Toxicology and Chemistry, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 7, 2025

Abstract In silico methods are increasingly important in predicting the ecotoxicity of engineered nanomaterials (ENMs), encompassing both individual and mixture toxicity predictions. It is widely recognized that ENMs trigger oxidative stress effects by generating intracellular reactive oxygen species (ROS), serving as a key mechanism their cytotoxicity studies. However, existing still face significant challenges induced ENMs. Herein, we utilized laboratory-derived data machine learning to develop quantitative nanostructure-activity relationship (nano-QSAR) classification regression models, aiming predict five carbon (fullerene, graphene, graphene oxide, single-walled nanotubes, multi-walled nanotubes) binary mixtures on Scenedesmus obliquus cells. We constructed nano-QSAR models combining zeta potential (ζP) with C4.5 decision tree, support vector machine, artificial neural network, naive Bayes, K-nearest neighbor algorithms. Moreover, three integrating features including ζP, hydrodynamic diameter (DH), specific surface area (SSA) logistic regression, random forest, Adaboost The Accuracy, Recall, Precision harmonic mean Recall (F1-score) values these were all higher than 0.600, indicating an excellent performance distinguishing whether CNMs have generate ROS. addition, using DH, SSA descriptors, combined tree forest gradient boosting, algorithm, successfully four applicable application domains (all training testing points lie within 95% confidence intervals), goodness-of-fit (Rtrain2 ≥ 0.850), robustness (cross-validation R2 0.650) well predictive power (Rtest2 0.610). method developed would establish fundamental basis for more precise evaluations ecological risks posed materials from mechanistic standpoint.

Language: Английский

---

The application of chemical similarity measures in an unconventional modeling framework c-RASAR along with dimensionality reduction techniques to a representative hepatotoxicity dataset

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Sept. 6, 2024

With the exponential progress in field of cheminformatics, conventional modeling approaches have so far been to employ supervised and unsupervised machine learning (ML) deep models, utilizing standard molecular descriptors, which represent structural, physicochemical, electronic properties a particular compound. Deviating from approach, this investigation, we employed classification Read-Across Structure-Activity Relationship (c-RASAR), involves amalgamation concepts classification-based quantitative structure-activity relationship (QSAR) incorporate Read-Across-derived similarity error-based descriptors into statistical framework. ML models developed these RASAR use similarity-based information close source neighbors query We different algorithms on selected QSAR develop predictive for efficient prediction compounds' hepatotoxicity. The predictivity each was evaluated large number test set compounds. best-performing model also used screen true external data set. explainable AI (XAI) coupled with were interpret contributions best c-RASAR explain chemical diversity dataset. application various dimensionality reduction techniques like t-SNE UMAP ARKA framework showed usefulness over their ability group similar compounds, enhancing modelability dataset efficiently identifying activity cliffs. Furthermore, cliffs identified by observing nature compounds constituting nearest On comparing our simple linear previously reported using same derived US FDA Orange Book ( https://www.accessdata.fda.gov/scripts/cder/ob/index.cfm ), it observed that is simple, reproducible, transferable, highly predictive. performance LDA supersedes work. Therefore, present can be predict hepatotoxicity chemicals.

Language: Английский

---

Machine learning assisted classification RASAR modeling for the nephrotoxicity potential of a curated set of orally active drugs

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 4, 2025

We have adopted the classification Read-Across Structure–Activity Relationship (c-RASAR) approach in present study for machine-learning (ML)-based model development from a recently reported curated dataset of nephrotoxicity potential orally active drugs. initially developed ML models using nine different algorithms separately on topological descriptors (referred to as simply "descriptors" subsequent sections manuscript) and MACCS fingerprints "fingerprints" manuscript), thus generating 18 QSAR models. Using chemical spaces defined by modeling fingerprints, similarity error-based RASAR were computed, most discriminating used develop another set c-RASAR All 36 cross-validated 20 times with fivefold cross-validation strategy, their predictivity was checked test data. A multi-criteria decision-making strategy – Sum Ranking Differences (SRD) approach—was identify best-performing based robustness external validation parameters. This statistical analysis suggested that had an overall good performance, while also (LDA derived descriptors, MCC values 0.229 0.431 training sets, respectively). screen true data prepared known nephrotoxic compounds DrugBankDB, demonstrating predictivity.

Language: Английский

---

Web Server-based Deep Learning-Driven Predictive Models for Respiratory Toxicity of Environmental Chemicals: Mechanistic Insights and Interpretability

Journal of Hazardous Materials, Journal Year: 2025, Volume and Issue: 489, P. 137575 - 137575

Published: Feb. 10, 2025

Language: Английский

---

PBScreen: A Server for the High-Throughput Screening of Placental Barrier–Permeable Contaminants Based on Multifusion Deep Learning

Environmental Pollution, Journal Year: 2025, Volume and Issue: unknown, P. 125858 - 125858

Published: Feb. 1, 2025

Language: Английский

---

Explainable machine learning models enhance prediction of PFAS bioactivity using quantitative molecular surface analysis-derived representation

Water Research, Journal Year: 2025, Volume and Issue: unknown, P. 123500 - 123500

Published: March 1, 2025

Language: Английский

---

Modeling and Interpretability Study of the Structure–Activity Relationship for Multigeneration EGFR Inhibitors

ACS Omega, Journal Year: 2025, Volume and Issue: unknown

Published: March 14, 2025

The fourth-generation EGFR inhibitors targeting L858R/T790M/C797S mutations are in clinical trials mostly, and it is necessary to develop new inhibitors. In this study, an internal data set containing 2302 multitarget the wild type (83%) L858R (92%), L858R/T790M (96%), (60%) was collected. We established a structure-activity relationship model for predicting bioactivities of multigeneration by multitask deep neural network (MT-DNN). also constructed four single-task models on 1384 mutation support vector machine (SVM), random forest (RF), XGBoost (XGB), single-target (ST-DNN), respectively. MT-DNN significantly outperformed external 304 Furthermore, combined application SHAP/delta-SHAP value interpretability analysis offers rigorous structural information from global perspective. With methods, can mine core scaffold important fragments provide valuable perspective address resistant problem.

Language: Английский

---

Introduction to Machine Learning for Predictive Modeling I

Challenges and advances in computational chemistry and physics, Journal Year: 2025, Volume and Issue: unknown, P. 3 - 30

Published: Jan. 1, 2025

Language: Английский

---

Risk Assessment of Industrial Chemicals Towards Salmon Species Amalgamating QSAR, q-RASAR, and ARKA Framework

Toxicology Reports, Journal Year: 2025, Volume and Issue: unknown, P. 102017 - 102017

Published: April 1, 2025

Language: Английский

---