Comprehensive hepatotoxicity prediction: ensemble model integrating machine learning and deep learning

Frontiers in Pharmacology, Journal Year: 2024, Volume and Issue: 15

Published: Aug. 21, 2024

Background Chemicals may lead to acute liver injuries, posing a serious threat human health. Achieving the precise safety profile of compound is challenging due complex and expensive testing procedures. In silico approaches will aid in identifying potential risk drug candidates initial stage development thus mitigating developmental cost. Methods current studies, QSAR models were developed for hepatotoxicity predictions using ensemble strategy integrate machine learning (ML) deep (DL) algorithms various molecular features. A large dataset 2588 chemicals drugs was randomly divided into training (80%) test (20%) sets, followed by individual base diverse or based on three different kinds descriptors fingerprints. Feature selection employed proceed with model optimizations performance. Hybrid further utilized determine method best Results The voting classifier emerged as optimal model, achieving an excellent prediction accuracy 80.26%, AUC 82.84%, recall over 93% bagging stacking classifiers method. verified external set, internal 10-fold cross-validation, rigorous benchmark training, exhibiting much better reliability than published models. Conclusion proposed offers dependable assessment good performance regarding induce damage.

Language: Английский

---

High‐throughput detection of craniofacial defects in fluorescent zebrafish

Birth Defects Research, Journal Year: 2022, Volume and Issue: 115(3), P. 371 - 389

Published: Nov. 11, 2022

Abstract Losses and malformations of cranial neural crest cell (cNCC) derivatives are a hallmark several common brain face malformations. Nevertheless, the etiology these cNCC defects remains unknown for many cases, suggesting complex basis involving interactions between genetic and/or environmental factors. However, sheer number possible factors (thousands genes hundreds thousands toxicants) has hindered identification specific interactions. Here, we develop high‐throughput analysis that will enable faster multifactorial in genesis craniofacial defects. Zebrafish embryos expressing fluorescent marker cNCCs ( fli1:EGFP ) were exposed to pathway inhibitor standard or toxicant, resulting changes fluorescence measured using microplate reader approximate losses. Embryos Hedgehog piperonyl butoxide (PBO), standard, alcohol (ethanol) exhibited reduced at one day post fertilization, which corresponded with five days fertilization. Combining PBO co‐exposure paradigm synergistically fluorescence, demonstrating interaction. Using reporter transgenics, show plate assay is sensitive detecting alterations signaling, critical regulator development. We go on demonstrate this technique readily detects other important types, namely neurons. Together, findings novel vivo platform can predict developmental abnormalities high‐throughput.

Language: Английский

---

Thermal proteome profiling and machine learning modeling for dissecting chemical-protein interactions in environmental toxicology: A mini-review and perspectives

Critical Reviews in Environmental Science and Technology, Journal Year: 2024, Volume and Issue: 54(20), P. 1478 - 1500

Published: Feb. 26, 2024

High throughput in vitro assays for screening chemical hazards focus primarily on specific receptors that are linked with certain adverse outcome pathways, neglecting potential novel endpoints or pathways induced by emerging pollutants. Identifying target proteins interact pollutants contributes to finding molecular initiating events under the framework. Mass spectrometry-based thermal proteome profiling (TPP) have permitted uncovering binding targets of across whole proteome. Based principle thermally stabilized after chemicals, TPP differentiates protein determining soluble fraction remain stable heat stress. Thus, facilitates qualitative and quantitative measurements chemical-protein interactions (CPIs) without modifications structures immobilization proteins. In this mini-review, we introduced principles, development procedures TPP, summarized its applications identifying speculating toxicity environmental toxicological studies. Additionally, since CPIs using multiple chemicals could be labor- cost-intensive, machine learning-based modeling is a feasible alternative dissect due capability mine intrinsic properties CPIs. Therefore, recent learning models CPI prediction was reviewed. Lastly, envisioned prospects combining data prediction, possibility applying interpret phenotypes generated from multi-omics data, inform future research forecasting outcomes

Language: Английский

---

Underexplored Molecular Mechanisms of Toxicity

Published: June 3, 2024

Social biases may concentrate attention of researchers on a small number well-known molecules/mechanisms leaving others underexplored. In accordance with this view, central to mechanistic toxicology is narrow range molecular pathways that are assumed be involved in significant part responses toxicity. It unclear, however, if there other mechanisms which play important role toxicity events but overlooked by toxicology. To identify genes sensitive chemical exposures we used publicly available databases. First, data published chemical-gene interactions for 17,338 estimate their sensitivity exposures. Next, extracted publication numbers per gene 19,243 human from Find My Understudied Genes database. Threshold were applied both datasets using our algorithm chemically and insensitive well-studied underexplored genes. 1,110 highly GSEA Shiny GO analyses enriched biological categories. Metabolism fatty acids, amino glucose identified as These findings suggest future effort needed uncover the xenobiotics current epidemics metabolic diseases.

Language: Английский

---

Emerging investigator series: Nontargeted Screening of Aryl Hydrocarbon Receptor Agonists in Endangered Beluga Whales from the St. Lawrence Estuary: Beyond Legacy Contaminants

Environmental Science Processes & Impacts, Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 1, 2024

The elevated concentrations of organohalogen contaminants in the endangered St. Lawrence Estuary (SLE) belugas have prompted hypothesis that aryl hydrocarbon receptor (AhR) activity may be a contributor towards their potential adverse effects. While indirect associations between AhR and contaminant levels been reported SLE beluga tissues, was never directly measured. Using bioassays nontargeted analysis, this study contrasted agonist profiles pooled tissue extracts non-threatened Arctic belugas. Tissue exhibited significantly higher overall than belugas, with 2000s liver extract exerting blubber from same time period. Contrary to our expectations, well-known agonists detected by including polychlorinated biphenyls (PCBs), were only minor contributors observed activity. Instead, Tox21 suspect screening identified more polar chemicals, such as dyes natural indoles, contributors. Notably, product bromoindole selectively at high abundance further confirmed an agonist. These findings highlighted significance AhR-mediated toxicity pathway underscored importance novel agonists, particularly compounds, its induction.

Language: Английский

---