Development of a large-scale chemogenomics database to improve drug candidate selection and to understand mechanisms of chemical toxicity and action

Transcriptomic point of departure determination: a comparison of distribution-based and gene set-based approaches

A key step in assessing the potential human and environmental health risks of industrial and agricultural chemicals is to determine the toxicity point of departure (POD), which is the highest dose level that causes no adverse effect. Transcriptomic POD (tPOD) values have been suggested to accurately estimate toxicity POD values. One step in the most common approach for tPOD determination involves mapping genes to annotated gene sets, a process that might lead to substantial information loss particularly in species with poor gene annotation. Alternatively, methods that calculate tPOD values directly from the distribution of individual gene POD values omit this mapping step. Using rat transcriptome data for 79 molecules obtained from Open TG-GATEs (Toxicogenomics Project Genomics Assisted Toxicity Evaluation System), the hypothesis was tested that methods based on the distribution of all individual gene POD values will give a similar tPOD value to that obtained via the gene set-based method. Gene set-based tPOD values using four different gene set structures were compared to tPOD values from five different individual gene distribution methods. Results revealed a high tPOD concordance for all methods tested, especially for molecules with at least 300 dose-responsive probesets: for 90% of those molecules, the tPOD values from all methods were within 4-fold of each other. In addition, random gene sets based upon the structure of biological knowledge-derived gene sets produced tPOD values with a median absolute fold change of 1.3-1.4 when compared to the original biological knowledge-derived gene set counterparts, suggesting that little biological information is used in the gene set-based tPOD generation approach. These findings indicate using individual gene distributions to calculate a tPOD is a viable and parsimonious alternative to using gene sets. Importantly, individual gene distribution-based tPOD methods do not require knowledge of biological organization and can be applied to any species including those with poorly annotated gene sets.

Unveiling the Mechanisms of Bone Marrow Toxicity Induced by Lead Acetate Exposure

Lead (Pb), a widespread heavy metal, causes severe toxicity in human and animal organs (e.g., bone marrow), whereas the mechanisms of the bone marrow toxicity induced by Pb exposure are unclear. Hence, this study was designed to reveal the hub genes involved in Pb-induced bone marrow toxicity. GSE59894 dataset obtained from Gene Expression Omnibus (GEO) was composed of lead acetate (PbAc2)-treated and control bone marrow samples. Totally 120 and 85 differentially expressed genes (DEGs) were identified on the 1st day, while 153 and 157 DEGs on the 3rd day in the bone marrow treated with 200 and 600 mg/kg of PbAc2, respectively. Notably, a total of 28 and 32 overlapping DEGs were identified in the bone marrow on the 1st and 3rd day treated with PbAc2, respectively. Biological process analysis suggested that the common DEGs were primarily participated in cell differentiation, the response to drug, xenobiotic stimulus, and organic cyclic compound. Pathway analysis demonstrated that the overlapping DEGs were primarily linked to PI3K-Akt, TGF-β, MAPK, and osteoclast differentiation signaling pathways. Moreover, the hub genes, including PLD2, DAPK1, ALB, TNF, FOS, CDKN1A, and TGFB3, might contribute to PbAc2-induced bone marrow toxicity. Overall, our study offers an important insight into the molecular mechanisms of Pb-induced bone marrow toxicity.

Exploring DrugCentral: from molecular structures to clinical effects

DrugCentral, accessible at https://drugcentral.org , is an open-access online drug information repository. It covers over 4950 drugs, incorporating structural, physicochemical, and pharmacological details to support drug discovery, development, and repositioning. With around 20,000 bioactivity data points, manual curation enhances information from several major digital sources. Approximately 724 mechanism-of-action (MoA) targets offer updated drug target insights. The platform captures clinical data: over 14,300 on- and off-label uses, 27,000 contraindications, and around 340,000 adverse drug events from pharmacovigilance reports. DrugCentral encompasses information from molecular structures to marketed formulations, providing a comprehensive pharmaceutical reference. Users can easily navigate basic drug information and key features, making DrugCentral a versatile, unique resource. Furthermore, we present a use-case example where we utilize experimentally determined data from DrugCentral to support drug repurposing. A minimum activity threshold t should be considered against novel targets to repurpose a drug. Analyzing 1156 bioactivities for human MoA targets suggests a general threshold of 1 µM: t = 6 when expressed as - log[Activity(M)]). This applies to 87% of the drugs. Moreover, t can be refined empirically based on water solubility (S): t = 3 - logS, for logS < - 3. Alongside the drug repurposing classification scheme, which considers intellectual property rights, market exclusivity protections, and market accessibility, DrugCentral provides valuable data to prioritize candidates for drug repurposing programs efficiently.

Application of transfer learning to predict drug-induced human in vivo gene expression changes using rat in vitro and in vivo data

The liver is the primary site for the metabolism and detoxification of many compounds, including pharmaceuticals. Consequently, it is also the primary location for many adverse reactions. As the liver is not readily accessible for sampling in humans; rodent or cell line models are often used to evaluate potential toxic effects of a novel compound or candidate drug. However, relating the results of animal and in vitro studies to relevant clinical outcomes for the human in vivo situation still proves challenging. In this study, we incorporate principles of transfer learning within a deep artificial neural network allowing us to leverage the relative abundance of rat in vitro and in vivo exposure data from the Open TG-GATEs data set to train a model to predict the expected pattern of human in vivo gene expression following an exposure given measured human in vitro gene expression. We show that domain adaptation has been successfully achieved, with the rat and human in vitro data no longer being separable in the common latent space generated by the network. The network produces physiologically plausible predictions of human in vivo gene expression pattern following an exposure to a previously unseen compound. Moreover, we show the integration of the human in vitro data in the training of the domain adaptation network significantly improves the temporal accuracy of the predicted rat in vivo gene expression pattern following an exposure to a previously unseen compound. In this way, we demonstrate the improvements in prediction accuracy that can be achieved by combining data from distinct domains.

Advancing Computational Toxicology by Interpretable Machine Learning

Chemical toxicity evaluations for drugs, consumer products, and environmental chemicals have a critical impact on human health. Traditional animal models to evaluate chemical toxicity are expensive, time-consuming, and often fail to detect toxicants in humans. Computational toxicology is a promising alternative approach that utilizes machine learning (ML) and deep learning (DL) techniques to predict the toxicity potentials of chemicals. Although the applications of ML- and DL-based computational models in chemical toxicity predictions are attractive, many toxicity models are "black boxes" in nature and difficult to interpret by toxicologists, which hampers the chemical risk assessments using these models. The recent progress of interpretable ML (IML) in the computer science field meets this urgent need to unveil the underlying toxicity mechanisms and elucidate the domain knowledge of toxicity models. In this review, we focused on the applications of IML in computational toxicology, including toxicity feature data, model interpretation methods, use of knowledge base frameworks in IML development, and recent applications. The challenges and future directions of IML modeling in toxicology are also discussed. We hope this review can encourage efforts in developing interpretable models with new IML algorithms that can assist new chemical assessments by illustrating toxicity mechanisms in humans.

A generative adversarial network model alternative to animal studies for clinical pathology assessment

Animal studies are unavoidable in evaluating chemical and drug safety. Generative Adversarial Networks (GANs) can generate synthetic animal data by learning from the legacy animal study results, thus may serve as an alternative approach to assess untested chemicals. AnimalGAN, a GAN method to simulate 38 rat clinical pathology measures, was developed with significant robustness even for the drugs that vary significantly from these used during training, both in terms of chemical structure, drug class, and the year of FDA approval. AnimalGAN showed comparable results in hepatotoxicity assessment as using the real animal data and outperformed 12 conventional quantitative structure-activity relationship approaches. Using AnimalGAN, a virtual experiment of 100,000 rats ranked hepatotoxicity of three structurally similar drugs in a similar trend that has been observed in human population. AnimalGAN represented a significant step with artificial intelligence towards the global effort in replacement, reduction, and refinement (3Rs) of animal use.

Assessing Kidney Injury Induced by Mercuric Chloride in Guinea Pigs with In Vivo and In Vitro Experiments

Acute kidney injury, which is associated with high levels of morbidity and mortality, affects a significant number of individuals, and can be triggered by multiple factors, such as medications, exposure to toxic chemicals or other substances, disease, and trauma. Because the kidney is a critical organ, understanding and identifying early cellular or gene-level changes can provide a foundation for designing medical interventions. In our earlier work, we identified gene modules anchored to histopathology phenotypes associated with toxicant-induced liver and kidney injuries. Here, using in vivo and in vitro experiments, we assessed and validated these kidney injury-associated modules by analyzing gene expression data from the kidneys of male Hartley guinea pigs exposed to mercuric chloride. Using plasma creatinine levels and cell-viability assays as measures of the extent of renal dysfunction under in vivo and in vitro conditions, we performed an initial range-finding study to identify the appropriate doses and exposure times associated with mild and severe kidney injuries. We then monitored changes in kidney gene expression at the selected doses and time points post-toxicant exposure to characterize the mechanisms of kidney injury. Our injury module-based analysis revealed a dose-dependent activation of several phenotypic cellular processes associated with dilatation, necrosis, and fibrogenesis that were common across the experimental platforms and indicative of processes that initiate kidney damage. Furthermore, a comparison of activated injury modules between guinea pigs and rats indicated a strong correlation between the modules, highlighting their potential for cross-species translational studies.

Identifying multiscale translational safety biomarkers using a network-based systems approach

Animal testing is the current standard for drug and chemicals safety assessment, but hazards translation to human is uncertain. Human in vitro models can address the species translation but might not replicate in vivo complexity. Herein, we propose a network-based method addressing these translational multiscale problems that derives in vivo liver injury biomarkers applicable to in vitro human early safety screening. We applied weighted correlation network analysis (WGCNA) to a large rat liver transcriptomic dataset to obtain co-regulated gene clusters (modules). We identified modules statistically associated with liver pathologies, including a module enriched for ATF4-regulated genes as associated with the occurrence of hepatocellular single-cell necrosis, and as preserved in human liver in vitro models. Within the module, we identified TRIB3 and MTHFD2 as a novel candidate stress biomarkers, and developed and used BAC-eGFPHepG2 reporters in a compound screening, identifying compounds showing ATF4-dependent stress response and potential early safety signals.

Computational drug repurposing by exploiting large-scale gene expression data: Strategy, methods and applications

De novo drug development is an extremely complex, time-consuming and costly task. Urgent needs for therapies of various diseases have greatly accelerated searches for more effective drug development methods. Luckily, drug repurposing provides a new and effective perspective on disease treatment. Rapidly increased large-scale transcriptome data paints a detailed prospect of gene expression during disease onset and thus has received wide attention in the field of computational drug repurposing. However, how to efficiently mine transcriptome data and identify new indications for old drugs remains a critical challenge. This review discussed the irreplaceable role of transcriptome data in computational drug repurposing and summarized some representative databases, tools and strategies. More importantly, it proposed a practical guideline through establishing the correspondence between three gene expression data types and five strategies, which would facilitate researchers to adopt appropriate strategies to deeply mine large-scale transcriptome data and discover more effective therapies.

Linking chemicals, genes and morphological perturbations to diseases

The progress in image-based high-content screening technology has facilitated high-throughput phenotypic profiling notably the quantification of cell morphology perturbation by chemicals. However, understanding the mechanism of action of a chemical and linking it to cell morphology and phenotypes remains a challenge in drug discovery. In this study, we intended to integrate molecules that induced transcriptomic perturbations and cellular morphological changes into a biological network in order to assess chemical-phenotypic relationships in humans. Such a network was enriched with existing disease information to suggest molecular and cellular profiles leading to phenotypes. Two datasets were used for this study. Firstly, we used the "Cell Painting morphological profiling assay" dataset, composed of 30,000 compounds tested on human osteosarcoma cells (named U2OS). Secondly, we used the "L1000 mRNA profiling assay" dataset, a collection of transcriptional expression data from cultured human cells treated with approximately 20,000 bioactive small molecules from the Library of Integrated Network-based Cellular Signatures (LINCS). Furthermore, pathways, gene ontology terms and disease enrichments were performed on the transcriptomics data. Overall, our study makes it possible to develop a biological network combining chemical-gene-pathway-morphological perturbation and disease relationships. It contains an ensemble of 9989 chemicals, 732 significant morphological features and 12,328 genes. Through diverse examples, we demonstrated that some drugs shared similar genes, pathways and morphological profiles that, taken together, could help in deciphering chemical-phenotype observations.

Associations of peripheral blood DNA methylation and estimated monocyte proportion differences during infancy with toddler attachment style

Attachment is a motivational system promoting felt security to a caregiver resulting in a persistent internal working model of interpersonal behavior. Attachment styles are developed in early social environments and predict future health and development outcomes with potential biological signatures, such as epigenetic modifications like DNA methylation (DNAm). Thus, we hypothesized infant DNAm would associate with toddler attachment styles. An epigenome-wide association study (EWAS) of blood DNAm from 3-month-old infants was regressed onto children's attachment style from the Strange Situation Procedure at 22-months at multiple DNAm Cytosine-phosphate-Guanine (CpG) sites. The 26 identified CpGs associated with proinflammatory immune phenotypes and cognitive development. In post-hoc analyses, only maternal cognitive-growth fostering, encouraging intellectual exploration, contributed. For disorganized children, DNAm-derived cell-type proportions estimated higher monocytes -cells in immune responses hypothesized to increase with early adversity. Collectively, these findings suggested the potential biological embedding of both adverse and advantageous social environments as early as 3-months-old.

Toxicogenomics scoring system: TGSS, a novel integrated risk assessment model for chemical carcinogenicity prediction

BACKGROUND

Given the increasing exposure of humans to environmental chemicals and the limitations of conventional toxicity test, there is an urgent need to develop next-generation risk assessment methods.

OBJECTIVES

This study aims to establish a novel computational system named Toxicogenomics Scoring System (TGSS) to predict the carcinogenicity of chemicals coupling chemical-gene interactions with multiple cancer transcriptomic datasets.

METHODS

Chemical-related gene signatures were derived from chemical-gene interaction data from the Comparative Toxicogenomics Database (CTD). For each cancer type in TCGA, genes were ranked by their effects on tumorigenesis, which is based on the differential expression between tumor and normal samples. Next, we developed carcinogenicity scores (C-scores) using pre-ranked GSEA to quantify the correlation between chemical-related gene signatures and ranked gene lists. Then we established TGSS by systematically evaluating the C-scores in multiple chemical-tumor pairs. Furthermore, we examined the performance of our approach by ROC curves or prognostic analyses in TCGA and multiple independent cancer cohorts.

RESULTS

Forty-six environmental chemicals were finally included in the study. C-score was calculated for each chemical-tumor pair. The C-scores of IARC Group 3 chemicals were significantly lower than those of chemicals in Group 1 (P-value = 0.02) and Group 2 (P-values = 7.49 ×10^-5). ROC curves analysis indicated that C-score could distinguish "high-risk chemicals" from the other compounds (AUC = 0.67) with a specificity and sensitivity of 0.86 and 0.57. The results of survival analysis were also in line with the assessed carcinogenicity in TGSS for the chemicals in Group 1. Finally, consistent results were further validated in independent cancer cohorts.

CONCLUSION

TGSS highlighted the great potential of integrating chemical-gene interactions with gene-cancer relationships to predict the carcinogenic risk of chemicals, which would be valuable for systems toxicology.

Connecting omics signatures and revealing biological mechanisms with iLINCS

There are only a few platforms that integrate multiple omics data types, bioinformatics tools, and interfaces for integrative analyses and visualization that do not require programming skills. Here we present iLINCS ( http://ilincs.org ), an integrative web-based platform for analysis of omics data and signatures of cellular perturbations. The platform facilitates mining and re-analysis of the large collection of omics datasets (>34,000), pre-computed signatures (>200,000), and their connections, as well as the analysis of user-submitted omics signatures of diseases and cellular perturbations. iLINCS analysis workflows integrate vast omics data resources and a range of analytics and interactive visualization tools into a comprehensive platform for analysis of omics signatures. iLINCS user-friendly interfaces enable execution of sophisticated analyses of omics signatures, mechanism of action analysis, and signature-driven drug repositioning. We illustrate the utility of iLINCS with three use cases involving analysis of cancer proteogenomic signatures, COVID 19 transcriptomic signatures and mTOR signaling.

CREB is a potential marker associated with drug-induced liver injury: Identification and validation through transcriptome database analysis

Drug-induced liver injury (DILI) is the main cause of failure in drug development and postapproval withdrawal. Although toxicogenomic techniques provide an unprecedented opportunity for mechanistic assessment and biomarker discovery, they are not suitable for the screening of large numbers of exploratory compounds in early drug discovery. Using a comprehensive analysis of toxicogenomics (TGx) data, we aimed to find DILI-relevant transcription factors (TFs) that could be incorporated into a reporter gene assay system. Gene set enrichment analysis (GSEA) of the Open TG-GATEs dataset highlighted 4 DILI-relevant TFs, including CREB, NRF2, ELK-1, and E2F. Using ten drugs with already assigned idiosyncratic toxicity (IDT) risks, reporter gene assays were conducted in HepG2 cells in the presence of the S9 mix. There were weak correlations between NRF2 activity and IDT risk, whereas strong correlations were observed between CREB activity and IDT risk. In addition, CREB activation associated with 3 Withdrawn/Black box Warning drugs was reversed by pretreatment with a PKA inhibitor. Collectively, we suggest that CREB might be a sensitive biomarker for DILI prediction, and its response to stress induced by high-risk drugs might be primarily regulated by the PKA/CREB signaling pathway.

Grouping of UVCB substances with dose-response transcriptomics data from human cell-based assays

The application of in vitro biological assays as new approach methodologies (NAMs) to support grouping of UVCB (unknown or variable composition, complex reaction products, and biological materials) substances has recently been demonstrated. In addition to cell-based phenotyping as NAMs, in vitro transcriptomic profiling is used to gain deeper mechanistic understanding of biological responses to chemicals and to support grouping and read-across. However, the value of gene expression profiling for characterizing complex substances like UVCBs has not been explored. Using 141 petroleum substance extracts, we performed dose-response transcriptomic profiling in human induced pluripotent stem cell (iPSC)-derived hepatocytes, cardiomyocytes, neurons, and endothelial cells, as well as cell lines MCF7 and A375. The goal was to determine whether transcriptomic data can be used to group these UVCBs and to further characterize the molecular basis for in vitro biological responses. We found distinct transcriptional responses for petroleum substances by manufacturing class. Pathway enrichment informed interpretation of effects of substances and UVCB petroleum-class. Transcriptional activity was strongly correlated with concentration of polycyclic aromatic compounds (PAC), especially in iPSC-derived hepatocytes. Supervised analysis using transcriptomics, alone or in combination with bioactivity data collected on these same substances/cells, suggest that transcriptomics data provide useful mechanistic information, but only modest additional value for grouping. Overall, these results further demonstrate the value of NAMs for grouping of UVCBs, identify informative cell lines, and provide data that could be used for justifying selection of substances for further testing that may be required for registration.

Machine learning and artificial intelligence in toxicological sciences

Machine learning and artificial intelligence approaches have revolutionized multiple disciplines, including toxicology. This review summarizes representative recent applications of machine learning and artificial intelligence approaches in different areas of toxicology, including physiologically based pharmacokinetic (PBPK) modeling, quantitative structure-activity relationship modeling for toxicity prediction, adverse outcome pathway analysis, high-throughput screening, toxicogenomics, big data and toxicological databases. By leveraging machine learning and artificial intelligence approaches, now it is possible to develop PBPK models for hundreds of chemicals efficiently, to create in silico models to predict toxicity for a large number of chemicals with similar accuracies compared to in vivo animal experiments, and to analyze a large amount of different types of data (toxicogenomics, high-content image data, etc.) to generate new insights into toxicity mechanisms rapidly, which was impossible by manual approaches in the past. To continue advancing the field of toxicological sciences, several challenges should be considered: (1) not all machine learning models are equally useful for a particular type of toxicology data, and thus it is important to test different methods to determine the optimal approach; (2) current toxicity prediction is mainly on bioactivity classification (yes/no), so additional studies are needed to predict the intensity of effect or dose-response relationship; (3) as more data become available, it is crucial to perform rigorous data quality check and develop infrastructure to store, share, analyze, evaluate, and manage big data; and (4) it is important to convert machine learning models to user-friendly interfaces to facilitate their applications by both computational and bench scientists.

Variations of SARS-CoV-2 in the Iranian population and candidate putative drug-like compounds to inhibit the mutated proteins

The first cases of the novel coronavirus, SARS-CoV-2, were detected in December 2019 in Wuhan, China. Nucleotide substitutions and mutations in the SARS-CoV-2 sequence can result in the evolution of the virus and its rapid spread across the world. Therefore, understanding genetic variants of SARS-CoV-2 and targeting the conserved elements responsible for viral replication have great benefits for detecting its infection sources and diagnosing and treating COVID-19. In this study, we used the SARS-CoV-2 sequence isolated from a 59-year-old man in Ardabil, Iran, in April 2020 and sequenced using Oxford Nanopore technology. A meta-analysis comparing the sequence under study with other sequences from Iran indicated long nucleotide insertions/deletions (indels) that code for NSP15, the NSP14-NSP10 complex, open reading frame ORF9b, and ORF1ab polyproteins. In addition, replicating the NSP8 protein in the study sequence is another topic that can affect viral replication. Then using the DNA structure of NSP8, NSP15, NSP14-NSP10 complex, and ORF1ab as a genetic target can help find drug-like compounds for COVID-19. Potential drug-like compounds reported in this study for their mechanism of action and interactions with SARS-CoV-2 genes using drug repurposing are resveratrol, erythromycin, chloramphenicol, indomethacin, ciclesonide, and PDE4 inhibitor. Ciclesonide appears to show the best results when docked with chosen viral proteins. Therefore, different proteins isolated from nucleotide mutations in the virus sequence can indicate distinct inducers for antibodies and are important in vaccine design.

Network approaches for modeling the effect of drugs and diseases

The network approach is quickly becoming a fundamental building block of computational methods aiming at elucidating the mechanism of action (MoA) and therapeutic effect of drugs. By modeling the effect of drugs and diseases on different biological networks, it is possible to better explain the interplay between disease perturbations and drug targets as well as how drug compounds induce favorable biological responses and/or adverse effects. Omics technologies have been extensively used to generate the data needed to study the mechanisms of action of drugs and diseases. These data are often exploited to define condition-specific networks and to study whether drugs can reverse disease perturbations. In this review, we describe network data mining algorithms that are commonly used to study drug’s MoA and to improve our understanding of the basis of chronic diseases. These methods can support fundamental stages of the drug development process, including the identification of putative drug targets, the in silico screening of drug compounds and drug combinations for the treatment of diseases. We also discuss recent studies using biological and omics-driven networks to search for possible repurposed FDA-approved drug treatments for SARS-CoV-2 infections (COVID-19).

Integrated Genomic and Bioinformatics Approaches to Identify Molecular Links between Endocrine Disruptors and Adverse Outcomes

Exposure to Endocrine Disrupting Chemicals (EDC) has been linked with several adverse outcomes. In this review, we examine EDCs that are pervasive in the environment and are of concern in the context of human, animal, and environmental health. We explore the consequences of EDC exposure on aquatic life, terrestrial animals, and humans. We focus on the exploitation of genomics technologies and in particular whole transcriptome sequencing. Genome-wide analyses using RNAseq provides snap shots of cellular, tissue and whole organism transcriptomes under normal physiological and EDC perturbed conditions. A global view of gene expression provides highly valuable information as it uncovers gene families or more specifically, pathways that are affected by EDC exposures, but also reveals those that are unaffected. Hypotheses about genes with unknown functions can also be formed by comparison of their expression levels with genes of known function. Risk assessment strategies leveraging genomic technologies and the development of toxicology databases are explored. Finally, we review how the Adverse Outcome Pathway (AOP) has exploited this high throughput data to provide a framework for toxicology studies.

OUP accepted manuscript

Purpose

Endocrine disruptors are a rising concern due to the wide array of health issues that it can cause. Although there are tools for mode of action (MoA)-based prediction of endocrine disruption (e.g. QSAR Toolbox and iSafeRat), none of them is based on toxicogenomics data. Here, we present EDTox, an R Shiny application enabling users to explore and use a computational method that we have recently published to identify and prioritize endocrine disrupting (ED) chemicals based on toxicogenomic data. The EDTox pipeline utilizes previously trained toxicogenomic-driven classifiers to make predictions on new untested compounds by using their molecular initiating events. Furthermore, the proposed R Shiny app allows users to extend the prediction systems by training and adding new classifiers based on new available toxicogenomic data. This functionality helps users to explore the ED potential of chemicals in new, untested exposure scenarios.

Availability and implementation

This tool is available as web application (www.edtox.fi) and stand-alone software on GitHub and Zenodo (https://doi.org/10.5281/zenodo.5817093).

Supplementary information

Supplementary data are available at Bioinformatics online.

Emerging Bioinformatics Methods and Resources in Drug Toxicology

Assessing the drug safety at an early stage of a drug discovery program is a critical issue. With the recent advances in molecular biology and genomic, massive amounts of generated and accumulated data by advanced experimental technologies such as RNA sequencing or proteomics start to be at the disposal of the scientific community. Innovative and adequate bioinformatic methods, tools, and protocols are required to analyze properly these diverse and extensive data sources with the aim to identify key features that are related to toxicity observations. Furthermore, the assessment of drug safety can be performed across multiple scales of complexity from molecular, cellular to phenotypic levels; therefore, the application of network science contributes to a better interpretation of the drug's exposure effect on human health. Here, we review databases containing toxicogenomics and chemical-phenotype information, as well as appropriated bioinformatics approaches that are currently used to analyze such data. Extension to others methods such as dose-responses, time-dependent processes, and text mining is also presented giving an overview of suitable tools available for a best practice of drug safety analysis.

Tox-GAN: An AI Approach Alternative to Animal Studies-a Case Study with Toxicogenomics

Animal studies are a critical component in biomedical research, pharmaceutical product development, and regulatory submissions. There is a worldwide effort in toxicology towards "reducing, refining and replacing" (3Rs) animal use. Here, we proposed a deep generative adversarial network (GAN)-based framework capable of deriving new animal results from existing animal studies without additional experiments. To prove the concept, we employed this Tox-GAN framework to generate both gene activities and expression profiles for multiple doses and treatment durations in toxicogenomics (TGx). Using the pre-existing rat liver TGx data from the Open TG-GATEs, we generated Tox-GAN transcriptomic profiles with high similarity (0.997 ± 0.002 in intensity and 0.740 ± 0.082 in fold change) to the corresponding real gene expression profiles. Consequently, Tox-GAN showed an outstanding performance in two critical TGx applications, gaining a molecular understanding of underlying toxicological mechanisms and gene expression-based biomarker development. For the former, over 87% agreement in Gene Ontology was found between Tox-GAN results and real gene expression data. For the latter, the concordance of biomarkers between real and generated data was high in both predictive performance and biomarker genes. We also demonstrated that the Tox-GAN models constructed with TG-GATEs data were capable of generating transcriptomic profiles reported in DrugMatrix. Finally, we demonstrated potential utility for Tox-GAN in aiding chemical-based read-across. To the best of our knowledge, the proposed Tox-GAN model is novel in its ability to generate in vivo transcriptomic profiles at different treatment conditions from chemical structures. Overall, Tox-GAN holds great promise for generating high-quality toxicogenomic profiles without animal experimentation.

CEBS update: curated toxicology database with enhanced tools for data integration

The Chemical Effects in Biological Systems database (CEBS) contains extensive toxicology study results and metadata from the Division of the National Toxicology Program (NTP) and other studies of environmental health interest. This resource grants public access to search and collate data from over 10 250 studies for 12 750 test articles (chemicals, environmental agents). CEBS has made considerable strides over the last 5 years to integrate growing internal data repositories into data warehouses and data marts to better serve the public with high quality curated datasets. This effort includes harmonizing legacy terms and metadata to current standards, mapping test articles to external identifiers, and aligning terms to OBO (Open Biological and Biomedical Ontology) Foundry ontologies. The data are made available through the CEBS Homepage (https://cebs.niehs.nih.gov/cebs/), guided search applications, flat files on FTP (file transfer protocol), and APIs (application programming interface) for user access and to provide a bridge for computational tools. The user interface is intuitive with a single search bar to query keywords related to study metadata, publications, and data availability. Results are consolidated to single pages for each test article with NTP conclusions, publications, individual studies, data collections, and links to related test articles and projects available together.

A toxicogenomic data space for system-level understanding and prediction of EDC-induced toxicity

Endocrine disrupting compounds (EDCs) are a persistent threat to humans and wildlife due to their ability to interfere with endocrine signaling pathways. Inspired by previous work to improve chemical hazard identification through the use of toxicogenomics data, we developed a genomic-oriented data space for profiling the molecular activity of EDCs in an in silico manner, and for creating predictive models that identify and prioritize EDCs. Predictive models of EDCs, derived from gene expression data from rats (in vivo and in vitro primary hepatocytes) and humans (in vitro primary hepatocytes and HepG2), achieve testing accuracy greater than 90%. Negative test sets indicate that known safer chemicals are not predicted as EDCs. The rat in vivo-based classifiers achieve accuracy greater than 75% when tested for invitro to in vivoextrapolation. This study reveals key metabolic pathways and genes affected by EDCs together with a set of predictive models that utilize these pathways to prioritize EDCs in dose/time dependent manner and to predict EDCevokedmetabolic diseases.

Results of the Seventh Scientific Workshop of ECCO: Precision Medicine in IBD-What, Why, and How

Many diseases that affect modern humans fall in the category of complex diseases, thus called because they result from a combination of multiple aetiological and pathogenic factors. Regardless of the organ or system affected, complex diseases present major challenges in diagnosis, classification, and management. Current forms of therapy are usually applied in an indiscriminate fashion based on clinical information, but even the most advanced drugs only benefit a limited number of patients and to a variable and unpredictable degree. This 'one measure does not fit all' situation has spurred the notion that therapy for complex disease should be tailored to individual patients or groups of patients, giving rise to the notion of 'precision medicine' [PM]. Inflammatory bowel disease [IBD] is a prototypical complex disease where the need for PM has become increasingly clear. This prompted the European Crohn's and Colitis Organisation to focus the Seventh Scientific Workshop on this emerging theme. The articles in this special issue of the Journal address the various complementary aspects of PM in IBD, including what PM is; why it is needed and how it can be used; how PM can contribute to prediction and prevention of IBD; how IBD PM can aid in prognosis and improve response to therapy; and the challenges and future directions of PM in IBD. This first article of this series is structured on three simple concepts [what, why, and how] and addresses the definition of PM, discusses the rationale for the need of PM in IBD, and outlines the methodology required to implement PM in IBD in a correct and clinically meaningful way.

Quantitative Transcriptional Biomarkers of Xenobiotic Receptor Activation in Rat Liver for the Early Assessment of Drug Safety Liabilities

The robust transcriptional plasticity of liver mediated through xenobiotic receptors underlies its ability to respond rapidly and effectively to diverse chemical stressors. Thus, drug-induced gene expression changes in liver serve not only as biomarkers of liver injury, but also as mechanistic sentinels of adaptation in metabolism, detoxification, and tissue protection from chemicals. Modern RNA sequencing methods offer an unmatched opportunity to quantitatively monitor these processes in parallel and to contextualize the spectrum of dose-dependent stress, adaptation, protection, and injury responses induced in liver by drug treatments. Using this approach, we profiled the transcriptional changes in rat liver following daily oral administration of 120 different compounds, many of which are known to be associated with clinical risk for drug-induced liver injury by diverse mechanisms. Clustering, correlation, and linear modeling analyses were used to identify and optimize coexpressed gene signatures modulated by drug treatment. Here, we specifically focused on prioritizing 9 key signatures for their pragmatic utility for routine monitoring in initial rat tolerability studies just prior to entering drug development. These signatures are associated with 5 canonical xenobiotic nuclear receptors (AHR, CAR, PXR, PPARα, ER), 3 mediators of reactive metabolite-mediated stress responses (NRF2, NRF1, P53), and 1 liver response following activation of the innate immune response. Comparing paradigm chemical inducers of each receptor to the other compounds surveyed enabled us to identify sets of optimized gene expression panels and associated scoring algorithms proposed as quantitative mechanistic biomarkers with high sensitivity, specificity, and quantitative accuracy. These findings were further qualified using public datasets, Open TG-GATEs and DrugMatrix, and internal development compounds. With broader collaboration and additional qualification, the quantitative toxicogenomic framework described here could inform candidate selection prior to committing to drug development, as well as complement and provide a deeper understanding of the conventional toxicology study endpoints used later in drug development.

Scoring functions for drug-effect similarity

MOTIVATION

The difficulty to find new drugs and bring them to the market has led to an increased interest to find new applications for known compounds. Biological samples from many disease contexts have been extensively profiled by transcriptomics, and, intuitively, this motivates to search for compounds with a reversing effect on the expression of characteristic disease genes. However, disease effects may be cell line-specific and also depend on other factors, such as genetics and environment. Transcription profile changes between healthy and diseased cells relate in complex ways to profile changes gathered from cell lines upon stimulation with a drug. Despite these differences, we expect that there will be some similarity in the gene regulatory networks at play in both situations. The challenge is to match transcriptomes for both diseases and drugs alike, even though the exact molecular pathology/pharmacogenomics may not be known.

RESULTS

We substitute the challenge to match a drug effect to a disease effect with the challenge to match a drug effect to the effect of the same drug at another concentration or in another cell line. This is welldefined, reproducible in vitro and in silico and extendable with external data. Based on the Connectivity Map (CMap) dataset, we combined 26 different similarity scores with six different heuristics to reduce the number of genes in the model. Such gene filters may also utilize external knowledge e.g. from biological networks. We found that no similarity score always outperforms all others for all drugs, but the Pearson correlation finds the same drug with the highest reliability. Results are improved by filtering for highly expressed genes and to a lesser degree for genes with large fold changes. Also a network-based reduction of contributing transcripts was beneficial, here implemented by the FocusHeuristics. We found no drop in prediction accuracy when reducing the whole transcriptome to the set of 1000 landmark genes of the CMap's successor project Library of Integrated Network-based Cellular Signatures. All source code to re-analyze and extend the CMap data, the source code of heuristics, filters and their evaluation are available to propel the development of new methods for drug repurposing.

AVAILABILITY

https://bitbucket.org/ibima/moldrugeffectsdb.

CONTACT

steffen.moeller@uni-rostock.de.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Briefings in Bioinformatics online.

Mexican Ganoderma Lucidum Extracts Decrease Lipogenesis Modulating Transcriptional Metabolic Networks and Gut Microbiota in C57BL/6 Mice Fed with a High-Cholesterol Diet

Prevention of hyperlipidemia and associated diseases is a health priority. Natural products, such as the medicinal mushroom Ganoderma lucidum (Gl), have demonstrated hypocholesterolemic, prebiotic and antidiabetic properties. However, the underlying transcriptomic mechanisms by which Gl exerts bioactivities are not completely understood. We report a comprehensive hepatic and renal transcriptome profiling of C57BL/6 mice under the consumption of a high-cholesterol diet and two standardized Gl extracts obtained from basidiocarps cultivated on conventional substrate (Gl-1) or substrate containing acetylsalicylic acid (ASA; Gl-2). We showed that Gl extracts modulate relevant metabolic pathways involving the restriction of lipid biosynthesis and the enrichment of lipid degradation and secretion. The Gl-2 extract exerts a major modulation over gene expression programs showing the highest similarity with simvastatin druggable-target-genes and these are enriched more in processes related to human obesity alterations in the liver. We further show a subset of Gl-modulated genes correlated with Lactobacillus enrichment and the reduction of circulating cholesterol-derived fats. Moreover, Gl extracts induce a significant decrease of macrophage lipid storage, which occurs concomitantly with the down-modulation of Fasn and Elovl6. Collectively, this evidence suggests a new link between Gl hypocholesterolemic and prebiotic activity, revealing thereby that standardized Mexican Gl extracts are a novel transcriptome modulator to prevent metabolic disorders associated with hypercholesterolemia.

signatureSearch: environment for gene expression signature searching and functional interpretation

signatureSearch is an R/Bioconductor package that integrates a suite of existing and novel algorithms into an analysis environment for gene expression signature (GES) searching combined with functional enrichment analysis (FEA) and visualization methods to facilitate the interpretation of the search results. In a typical GES search (GESS), a query GES is searched against a database of GESs obtained from large numbers of measurements, such as different genetic backgrounds, disease states and drug perturbations. Database matches sharing correlated signatures with the query indicate related cellular responses frequently governed by connected mechanisms, such as drugs mimicking the expression responses of a disease. To identify which processes are predominantly modulated in the GESS results, we developed specialized FEA methods combined with drug-target network visualization tools. The provided analysis tools are useful for studying the effects of genetic, chemical and environmental perturbations on biological systems, as well as searching single cell GES databases to identify novel network connections or cell types. The signatureSearch software is unique in that it provides access to an integrated environment for GESS/FEA routines that includes several novel search and enrichment methods, efficient data structures, and access to pre-built GES databases, and allowing users to work with custom databases.

TREAP: A New Topological Approach to Drug Target Inference

Over 50% of drugs fail in stage 3 clinical trials, many because of a poor understanding of the drug's mechanisms of action (MoA). A better comprehension of drug MoA will significantly improve research and development (R&D). Current proposed algorithms, such as ProTINA and DeMAND, can be overly complex. Additionally, they are unable to predict whether the drug-induced gene expression or the topology of the networks used to model gene regulation primarily impacts accurate drug target inference. In this work, we evaluate how network and gene expression data affect ProTINA's accuracy. We find that network topology predominantly determines the accuracy of ProTINA's predictions. We further show that the size of an interaction network and/or selecting cell-specific networks has a limited effect on accuracy. We then demonstrate that a specific network topology measure, betweenness, can be used to improve drug target prediction. Based on these results, we create a new algorithm, TREAP, that combines betweenness values and adjusted p-values for target inference. TREAP offers an alternative approach to drug target inference and is advantageous because it is not computationally demanding, provides easy-to-interpret results, and is often more accurate at predicting drug targets than current state-of-the-art approaches.

Adding Stochastic Negative Examples into Machine Learning Improves Molecular Bioactivity Prediction

Multitask deep neural networks learn to predict ligand-target binding by example, yet public pharmacological data sets are sparse, imbalanced, and approximate. We constructed two hold-out benchmarks to approximate temporal and drug-screening test scenarios, whose characteristics differ from a random split of conventional training data sets. We developed a pharmacological data set augmentation procedure, Stochastic Negative Addition (SNA), which randomly assigns untested molecule-target pairs as transient negative examples during training. Under the SNA procedure, drug-screening benchmark performance increases from R² = 0.1926 ± 0.0186 to 0.4269 ± 0.0272 (122%). This gain was accompanied by a modest decrease in the temporal benchmark (13%). SNA increases in drug-screening performance were consistent for classification and regression tasks and outperformed y-randomized controls. Our results highlight where data and feature uncertainty may be problematic and how leveraging uncertainty into training improves predictions of drug-target relationships.

A Novel Open Access Web Portal for Integrating Mechanistic and Toxicogenomic Study Results

Applying toxicogenomics to improving the safety profile of drug candidates and crop protection molecules is most useful when it identifies relevant biological and mechanistic information that highlights risks and informs risk mitigation strategies. Pathway-based approaches, such as gene set enrichment analysis, integrate toxicogenomic data with known biological process and pathways. Network methods help define unknown biological processes and offer data reduction advantages. Integrating the 2 approaches would improve interpretation of toxicogenomic information. Barriers to the routine application of these methods in genome-wide transcriptomic studies include a need for “hands-on” computer programming experience, the selection of 1 or more analysis methods (eg pathway analysis methods), the sensitivity of results to algorithm parameters, and challenges in linking differential gene expression to variation in safety outcomes. To facilitate adoption and reproducibility of gene expression analysis in safety studies, we have developed Collaborative Toxicogeomics, an open-access integrated web portal using the Django web framework. The software, developed with the Python programming language, is modular, extensible and implements “best-practice” methods in computational biology. New study results are compared with over 4000 rodent liver experiments from Drug Matrix and open TG-GATEs. A unique feature of the software is the ability to integrate clinical chemistry and histopathology-derived outcomes with results from gene expression studies, leading to relevant mechanistic conclusions. We describe its application by analyzing the effects of several toxicants on liver gene expression and exemplify application to predicting toxicity study outcomes upon chronic treatment from expression changes in acute-duration studies.

Use of deep learning methods to translate drug-induced gene expression changes from rat to human primary hepatocytes

In clinical trials, animal and cell line models are often used to evaluate the potential toxic effects of a novel compound or candidate drug before progressing to human trials. However, relating the results of animal and in vitro model exposures to relevant clinical outcomes in the human in vivo system still proves challenging, relying on often putative orthologs. In recent years, multiple studies have demonstrated that the repeated dose rodent bioassay, the current gold standard in the field, lacks sufficient sensitivity and specificity in predicting toxic effects of pharmaceuticals in humans. In this study, we evaluate the potential of deep learning techniques to translate the pattern of gene expression measured following an exposure in rodents to humans, circumventing the current reliance on orthologs, and also from in vitro to in vivo experimental designs. Of the applied deep learning architectures applied in this study the convolutional neural network (CNN) and a deep artificial neural network with bottleneck architecture significantly outperform classical machine learning techniques in predicting the time series of gene expression in primary human hepatocytes given a measured time series of gene expression from primary rat hepatocytes following exposure in vitro to a previously unseen compound across multiple toxicologically relevant gene sets. With a reduction in average mean absolute error across 76 genes that have been shown to be predictive for identifying carcinogenicity from 0.0172 for a random regression forest to 0.0166 for the CNN model (p < 0.05). These deep learning architecture also perform well when applied to predict time series of in vivo gene expression given measured time series of in vitro gene expression for rats.

A toxicogenomic approach to assess kidney injury induced by mercuric chloride in rats

Kidney injury caused by disease, trauma, environmental exposures, or drugs may result in decreased renal function, chronic kidney disease, or acute kidney failure. Diagnosis of kidney injury using serum creatinine levels, a common clinical test, only identifies renal dysfunction after the kidneys have undergone severe damage. Other indicators sensitive to kidney injury, such as the level of urine kidney injury molecule-1 (KIM-1), lack the ability to differentiate between injury phenotypes. To address early detection as well as detailed categorization of kidney-injury phenotypes in preclinical animal or cellular studies, we previously identified eight sets (modules) of co-expressed genes uniquely associated with kidney histopathology. Here, we used mercuric chloride (HgCl₂)-a model nephrotoxicant-to chemically induce kidney injuries as monitored by KIM-1 levels in Sprague Dawley rats at two doses (0.25 or 0.50 mg/kg) and two exposure lengths (10 or 34 h). We collected whole transcriptome RNA-seq data derived from five animals at each dose and time point to perform a toxicogenomics analysis. Consistent with documented injury phenotypes for HgCl₂ toxicity, our kidney-injury-module approach identified the onset of necrosis and dilation as early as 10 h after a dose of 0.50 mg/kg that produced only mild injury as judged by urinary KIM-1 excretion. The results of these animal studies highlight the potential of the kidney-injury-module approach to provide a sensitive and histopathology-specific readout of renal toxicity.

Identification of Time-Invariant Biomarkers for Non-Genotoxic Hepatocarcinogen Assessment

Non-genotoxic hepatocarcinogens (NGHCs) can only be confirmed by 2-year rodent studies. Toxicogenomics (TGx) approaches using gene expression profiles from short-term animal studies could enable early assessment of NGHCs. However, high variance in the modulation of the genes had been noted among exposure styles and datasets. Expanding from our previous strategy in identifying consensus biomarkers in multiple experiments, we aimed to identify time-invariant biomarkers for NGHCs in short-term exposure styles and validate their applicability to long-term exposure styles. In this study, nine time-invariant biomarkers, namely A2m, Akr7a3, Aqp7, Ca3, Cdc2a, Cdkn3, Cyp2c11, Ntf3, and Sds, were identified from four large-scale microarray datasets. Machine learning techniques were subsequently employed to assess the prediction performance of the biomarkers. The biomarker set along with the Random Forest models gave the highest median area under the receiver operating characteristic curve (AUC) of 0.824 and a low interquartile range (IQR) variance of 0.036 based on a leave-one-out cross-validation. The application of the models to the external validation datasets achieved high AUC values of greater than or equal to 0.857. Enrichment analysis of the biomarkers inferred the involvement of chronic inflammatory diseases such as liver cirrhosis, fibrosis, and hepatocellular carcinoma in NGHCs. The time-invariant biomarkers provided a robust alternative for NGHC prediction.

Drug vector representation: a tool for drug similarity analysis

DrugMatrix is a valuable toxicogenomic dataset, which provides in vivo transcriptome data corresponding to hundreds of chemical drugs. However, the relationships between drugs and how those drugs affect the biological process are still unknown. The high dimensionality of the microarray data hinders its application. The aims of this study are to (1) represent the transcriptome data by lower-dimensional vectors, (2) compare drug similarity, (3) represent drug combinations by adding vectors and (4) infer drug mechanism of action (MoA) and genotoxicity features. We borrowed the latent semantic analysis (LSA) technique from natural language processing to represent treatments (drugs with multiple concentrations and time points) by dense vectors, each dimension of which is an orthogonal biological feature. The gProfiler enrichment tool was used for the 100-dimensional vector feature annotation. The similarity between treatments vectors was calculated by the cosine function. Adding vectors may represent drug combinations, treatment times or treatment doses that are not presented in the original data. Drug-drug interaction pairs had a higher similarity than random drug pairs in the hepatocyte data. The vector features helped to reveal the MoA. Differential feature expression was also implicated for genotoxic and non-genotoxic carcinogens. An easy-to-use Web tool was developed by Shiny Web application framework for the exploration of treatment similarities and drug combinations (https://bioinformatics.fafu.edu.cn/drugmatrix/). We represented treatments by vectors and provided a tool that is useful for hypothesis generation in toxicogenomic, such as drug similarity, drug repurposing, combination therapy and MoA.

Predict New Therapeutic Drugs for Hepatocellular Carcinoma Based on Gene Mutation and Expression

Hepatocellular carcinoma (HCC) is the fourth most common primary liver tumor and is an important medical problem worldwide. However, the use of current therapies for HCC is no possible to be cured, and despite numerous attempts and clinical trials, there are not so many approved targeted treatments for HCC. So, it is necessary to identify additional treatment strategies to prevent the growth of HCC tumors. We are looking for a systematic drug repositioning bioinformatics method to identify new drug candidates for the treatment of HCC, which considers not only aberrant genomic information, but also the changes of transcriptional landscapes. First, we screen the collection of HCC feature genes, i.e., kernel genes, which frequently mutated in most samples of HCC based on human mutation data. Then, the gene expression data of HCC in TCGA are combined to classify the kernel genes of HCC. Finally, the therapeutic score (TS) of each drug is calculated based on the kolmogorov-smirnov statistical method. Using this strategy, we identify five drugs that associated with HCC, including three drugs that could treat HCC and two drugs that might have side-effect on HCC. In addition, we also make Connectivity Map (CMap) profiles similarity analysis and KEGG enrichment analysis on drug targets. All these findings suggest that our approach is effective for accurate predicting novel therapeutic options for HCC and easily to be extended to other tumors.

Predicting drug-induced hepatotoxicity based on biological feature maps and diverse classification strategies

Identifying hepatotoxicity as early as possible is significant in drug development. In this study, we developed a drug-induced hepatotoxicity prediction model taking account of both the biological context and the computational efficacy based on toxicogenomics data. Specifically, we proposed a novel gene selection algorithm considering gene's participation, named BioCB, to choose the discriminative genes and make more efficient prediction. Then instead of using the raw gene expression levels to characterize each drug, we developed a two-dimensional biological process feature pattern map to represent each drug. Then we employed two strategies to handle the maps and identify the hepatotoxicity, the direct use of maps, named Two-dim branch, and vectorization of maps, named One-dim branch. The two strategies subsequently used the deep convolutional neural networks and LightGBM as predictors, respectively. Additionally, we here for the first time proposed a stacked vectorized gene matrix, which was more predictive than the raw gene matrix. Results validated on both in vivo and in vitro data from two public data sets, the TG-GATES and DrugMatrix, show that the proposed One-dim branch outperforms the deep framework, the Two-dim branch, and has achieved high accuracy and efficiency. The implementation of the proposed method is available at https://github.com/RanSuLab/Hepatotoxicity.

Drug repurposing to improve treatment of rheumatic autoimmune inflammatory diseases

The past century has been characterized by intensive efforts, within both academia and the pharmaceutical industry, to introduce new treatments to individuals with rheumatic autoimmune inflammatory diseases (RAIDs), often by 'borrowing' treatments already employed in one RAID or previously used in an entirely different disease, a concept known as drug repurposing. However, despite sharing some clinical manifestations and immune dysregulation, disease pathogenesis and phenotype vary greatly among RAIDs, and limited understanding of their aetiology has made repurposing drugs for RAIDs challenging. Nevertheless, the past century has been characterized by different 'waves' of repurposing. Early drug repurposing occurred in academia and was based on serendipitous observations or perceived disease similarity, often driven by the availability and popularity of drug classes. Since the 1990s, most biologic therapies have been developed for one or several RAIDs and then tested among the others, with varying levels of success. The past two decades have seen data-driven repurposing characterized by signature-based approaches that rely on molecular biology and genomics. Additionally, many data-driven strategies employ computational modelling and machine learning to integrate multiple sources of data. Together, these repurposing periods have led to advances in the treatment for many RAIDs.

Assessing Chemical-Induced Liver Injury In Vivo From In Vitro Gene Expression Data in the Rat: The Case of Thioacetamide Toxicity

Consumers are exposed to thousands of chemicals with potentially adverse health effects. However, these chemicals will never be tested for toxicity because of the immense resources needed for animal-based (in vivo) toxicological studies. Today, there are no viable in vitro alternatives to these types of animal studies. To develop an in vitro approach, we investigated whether we could predict in vivo organ injuries in rats with the use of RNA-seq data acquired from tissues early in the development of toxicant-induced injury, by comparing gene expression data from RNA isolated from these rat tissues with those obtained from in vitro exposure of primary liver and kidney cells. We collected RNA-seq data from the liver and kidney tissues of Sprague-Dawley rats 8 or 24 h after exposing them to vehicle (control), low (25 mg/kg), or high (100 mg/kg) doses of thioacetamide, a known liver toxicant that promotes fibrosis; we used these doses and exposure times to cause only mild toxicant-induced injury. For the in vitro study, we treated two cell types from Sprague-Dawley rats, primary hepatocytes (vehicle; low, 0.025 mM; or high, 0.125 mM dose), and renal tube epithelial cells (vehicle; low, 0.125 mM; or high, 0.500 mM) dose) with the thioacetamide metabolite, thioacetamide-S-oxide, selecting in vitro doses and exposure times to recreate the early-stage toxicant-induced injury model that we achieved in vivo. RNA-seq data were collected 9 or 24 h after application of vehicle or thioacetamide-S-oxide. We found that our modular approach for the analysis of gene expression data derived from in vivo RNA-seq strongly correlated (R² > 0.6) with the in vitro results at two different dose levels of thioacetamide/thioacetamide-S-oxide after 24 h of exposure. The top-ranked liver injury modules in vitro correctly identified the ensuing development of liver fibrosis.

Connectivity Mapping: Methods and Applications

Connectivity mapping resources consist of signatures representing changes in cellular state following systematic small-molecule, disease, gene, or other form of perturbations. Such resources enable the characterization of signatures from novel perturbations based on similarity; provide a global view of the space of many themed perturbations; and allow the ability to predict cellular, tissue, and organismal phenotypes for perturbagens. A signature search engine enables hypothesis generation by finding connections between query signatures and the database of signatures. This framework has been used to identify connections between small molecules and their targets, to discover cell-specific responses to perturbations and ways to reverse disease expression states with small molecules, and to predict small-molecule mimickers for existing drugs. This review provides a historical perspective and the current state of connectivity mapping resources with a focus on both methodology and community implementations.

The Carcinogenome Project: In Vitro Gene Expression Profiling of Chemical Perturbations to Predict Long-Term Carcinogenicity

BACKGROUND

Most chemicals in commerce have not been evaluated for their carcinogenic potential. The de facto gold-standard approach to carcinogen testing adopts the 2-y rodent bioassay, a time-consuming and costly procedure. High-throughput in vitro assays are a promising alternative for addressing the limitations in carcinogen screening.

OBJECTIVES

We developed a screening process for predicting chemical carcinogenicity and genotoxicity and characterizing modes of actions (MoAs) using in vitro gene expression assays.

METHODS

We generated a large toxicogenomics resource comprising [Formula: see text] expression profiles corresponding to 330 chemicals profiled in HepG2 (human hepatocellular carcinoma cell line) at multiple doses and replicates. Predictive models of carcinogenicity and genotoxicity were built using a random forest classifier. Differential pathway enrichment analysis was performed to identify pathways associated with carcinogen exposure. Signatures of carcinogenicity and genotoxicity were compared with external sources, including Drugmatrix and the Connectivity Map.

RESULTS

Among profiles with sufficient bioactivity, our classifiers achieved 72.2% Area Under the ROC Curve (AUC) for predicting carcinogenicity and 82.3% AUC for predicting genotoxicity. Chemical bioactivity, as measured by the strength and reproducibility of the transcriptional response, was not significantly associated with long-term carcinogenicity in doses up to [Formula: see text]. However, sufficient bioactivity was necessary for a chemical to be used for prediction of carcinogenicity. Pathway enrichment analysis revealed pathways consistent with known pathways that drive cancer, including DNA damage and repair. The data is available at https://clue.io/CRCGN_ABC , and a portal for query and visualization of the results is accessible at https://carcinogenome.org .

DISCUSSION

We demonstrated an in vitro screening approach using gene expression profiling to predict carcinogenicity and infer MoAs of chemical perturbations. https://doi.org/10.1289/EHP3986.

Advancing Computational Toxicology in the Big Data Era by Artificial Intelligence: Data-Driven and Mechanism-Driven Modeling for Chemical Toxicity

In 2016, the Frank R. Lautenberg Chemical Safety for the 21st Century Act became the first US legislation to advance chemical safety evaluations by utilizing novel testing approaches that reduce the testing of vertebrate animals. Central to this mission is the advancement of computational toxicology and artificial intelligence approaches to implementing innovative testing methods. In the current big data era, the terms volume (amount of data), velocity (growth of data), and variety (the diversity of sources) have been used to characterize the currently available chemical, in vitro, and in vivo data for toxicity modeling purposes. Furthermore, as suggested by various scientists, the variability (internal consistency or lack thereof) of publicly available data pools, such as PubChem, also presents significant computational challenges. The development of novel artificial intelligence approaches based on public massive toxicity data is urgently needed to generate new predictive models for chemical toxicity evaluations and make the developed models applicable as alternatives for evaluating untested compounds. In this procedure, traditional approaches (e.g., QSAR) purely based on chemical structures have been replaced by newly designed data-driven and mechanism-driven modeling. The resulting models realize the concept of adverse outcome pathway (AOP), which can not only directly evaluate toxicity potentials of new compounds, but also illustrate relevant toxicity mechanisms. The recent advancement of computational toxicology in the big data era has paved the road to future toxicity testing, which will significantly impact on the public health.

Comprehensive 3D-QSAR Model Predicts Binding Affinity of Structurally Diverse Sigma 1 Receptor Ligands

The Sigma 1 Receptor (S1R) has attracted intense interest as a pharmaceutical target for various therapeutic indications, including the treatment of neuropathic pain and the potentiation of opioid analgesia. Efforts by drug developers to rationally design S1R antagonists have been spurred recently by the 2016 publication of the high-resolution X-ray crystal structure of the ligand-bound human S1R. Until now, however, the absence in the published literature of a single, large-scale, and comprehensive quantitative structure-activity relationship (QSAR) model that encompasses a structurally diverse collection of S1R ligands has impaired rapid progress. To our best knowledge, the present study represents the first report of a statistically robust and highly predictive 3D-QSAR model (R2 = 0.92, Q2 = 0.62, Rpred2 = 0.81) based on the X-ray crystal structure of human S1R and constructed from a pooled compilation of 180 S1R antagonists that encompass five structurally diverse chemical families investigated using identical experimental protocols. Best practices, as recommended by the Organization for Economic Cooperation and Development (OECD: http://www.oecd.org/ ), were adopted for pooling data from disparate sources and for QSAR model development and both internal and external model validation. The practical utility of the final 3D-QSAR model was tested by virtual screening of the DrugBank database of FDA approved drugs supplemented by eight reported S1R antagonists. Among the top-ranked 40 DrugBank hits, four approved drugs which were previously unknown as S1R antagonists were tested using in vitro radiolabeled human S1R binding assays. Of these, two drugs (diphenhydramine and phenyltoloxamine) exhibited potent S1R binding affinity with Ki = 58 nM and 160 nM, respectively. As diphenhydramine is approved as an antiallergic, and phenyltoloxamine as an analgesic and sedative, each of these compounds represents a viable starting point for a drug discovery campaign aimed at the development of novel S1R antagonists for a wide range of therapeutic indications.

Application of the Transcriptional Disease Signature (TDSs) to Screen Melanoma-Effective Compounds in a Small Fish Model

Cell culture and protein target-based compound screening strategies, though broadly utilized in selecting candidate compounds, often fail to eliminate candidate compounds with non-target effects and/or safety concerns until late in the drug developmental process. Phenotype screening using intact research animals is attractive because it can help identify small molecule candidate compounds that have a high probability of proceeding to clinical use. Most FDA approved, first-in-class small molecules were identified from phenotypic screening. However, phenotypic screening using rodent models is labor intensive, low-throughput, and very expensive. As a novel alternative for small molecule screening, we have been developing gene expression disease profiles, termed the Transcriptional Disease Signature (TDS), as readout of small molecule screens for therapeutic molecules. In this concept, compounds that can reverse, or otherwise affect known disease-associated gene expression patterns in whole animals may be rapidly identified for more detailed downstream direct testing of their efficacy and mode of action. To establish proof of concept for this screening strategy, we employed a transgenic strain of a small aquarium fish, medaka (Oryzias latipes), that overexpresses the malignant melanoma driver gene xmrk, a mutant egfr gene, that is driven by a pigment cell-specific mitf promoter. In this model, melanoma develops with 100% penetrance. Using the transgenic medaka malignant melanoma model, we established a screening system that employs the NanoString nCounter platform to quantify gene expression within custom sets of TDS gene targets that we had previously shown to exhibit differential transcription among xmrk-transgenic and wild-type medaka. Compound-modulated gene expression was identified using an internet-accessible custom-built data processing pipeline. The effect of a given drug on the entire TDS profile was estimated by comparing compound-modulated genes in the TDS using an activation Z-score and Kolmogorov-Smirnov statistics. TDS gene probes were designed that target common signaling pathways that include proliferation, development, toxicity, immune function, metabolism and detoxification. These pathways may be utilized to evaluate candidate compounds for potential favorable, or unfavorable, effects on melanoma-associated gene expression. Here we present the logistics of using medaka to screen compounds, as well as, the development of a user-friendly NanoString data analysis pipeline to support feasibility of this novel TDS drug-screening strategy.

Comparison of RNA-Seq and Microarray Gene Expression Platforms for the Toxicogenomic Evaluation of Liver From Short-Term Rat Toxicity Studies

Gene expression profiling is a useful tool to predict and interrogate mechanisms of toxicity. RNA-Seq technology has emerged as an attractive alternative to traditional microarray platforms for conducting transcriptional profiling. The objective of this work was to compare both transcriptomic platforms to determine whether RNA-Seq offered significant advantages over microarrays for toxicogenomic studies. RNA samples from the livers of rats treated for 5 days with five tool hepatotoxicants (α-naphthylisothiocyanate/ANIT, carbon tetrachloride/CCl₄, methylenedianiline/MDA, acetaminophen/APAP, and diclofenac/DCLF) were analyzed with both gene expression platforms (RNA-Seq and microarray). Data were compared to determine any potential added scientific (i.e., better biological or toxicological insight) value offered by RNA-Seq compared to microarrays. RNA-Seq identified more differentially expressed protein-coding genes and provided a wider quantitative range of expression level changes when compared to microarrays. Both platforms identified a larger number of differentially expressed genes (DEGs) in livers of rats treated with ANIT, MDA, and CCl₄ compared to APAP and DCLF, in agreement with the severity of histopathological findings. Approximately 78% of DEGs identified with microarrays overlapped with RNA-Seq data, with a Spearman’s correlation of 0.7 to 0.83. Consistent with the mechanisms of toxicity of ANIT, APAP, MDA and CCl₄, both platforms identified dysregulation of liver relevant pathways such as Nrf2, cholesterol biosynthesis, eiF2, hepatic cholestasis, glutathione and LPS/IL-1 mediated RXR inhibition. RNA-Seq data showed additional DEGs that not only significantly enriched these pathways, but also suggested modulation of additional liver relevant pathways. In addition, RNA-Seq enabled the identification of non-coding DEGs that offer a potential for improved mechanistic clarity. Overall, these results indicate that RNA-Seq is an acceptable alternative platform to microarrays for rat toxicogenomic studies with several advantages. Because of its wider dynamic range as well as its ability to identify a larger number of DEGs, RNA-Seq may generate more insight into mechanisms of toxicity. However, more extensive reference data will be necessary to fully leverage these additional RNA-Seq data, especially for non-coding sequences.

Literature optimized integration of gene expression for organ-specific evaluation of toxicogenomics datasets

The study of drug toxicity in human organs is complicated by their complex inter-relations and by the obvious difficulty to testing drug effects on biologically relevant material. Animal models and human cell cultures offer alternatives for systematic and large-scale profiling of drug effects on gene expression level, as typically found in the so-called toxicogenomics datasets. However, the complexity of these data, which includes variable drug doses, time points, and experimental setups, makes it difficult to choose and integrate the data, and to evaluate the appropriateness of one or another model system to study drug toxicity (of particular drugs) of particular human organs. Here, we define a protocol to integrate drug-wise rankings of gene expression changes in toxicogenomics data, which we apply to the TG-GATEs dataset, to prioritize genes for association to drug toxicity in liver or kidney. Contrast of the results with sets of known human genes associated to drug toxicity in the literature allows to compare different rank aggregation approaches for the task at hand. Collectively, ranks from multiple models point to genes not previously associated to toxicity, notably, the PCNA clamp associated factor (PCLAF), and genes regulated by the master regulator of the antioxidant response NFE2L2, such as NQO1 and SRXN1. In addition, comparing gene ranks from different models allowed us to evaluate striking differences in terms of toxicity-associated genes between human and rat hepatocytes or between rat liver and rat hepatocytes. We interpret these results to point to the different molecular functions associated to organ toxicity that are best described by each model. We conclude that the expected production of toxicogenomics panels with larger numbers of drugs and models, in combination with the ongoing increase of the experimental literature in organ toxicity, will lead to increasingly better associations of genes for organism toxicity.

Large scale comparison of QSAR and conformal prediction methods and their applications in drug discovery

Structure–activity relationship modelling is frequently used in the early stage of drug discovery to assess the activity of a compound on one or several targets, and can also be used to assess the interaction of compounds with liability targets. QSAR models have been used for these and related applications over many years, with good success. Conformal prediction is a relatively new QSAR approach that provides information on the certainty of a prediction, and so helps in decision-making. However, it is not always clear how best to make use of this additional information. In this article, we describe a case study that directly compares conformal prediction with traditional QSAR methods for large-scale predictions of target-ligand binding. The ChEMBL database was used to extract a data set comprising data from 550 human protein targets with different bioactivity profiles. For each target, a QSAR model and a conformal predictor were trained and their results compared. The models were then evaluated on new data published since the original models were built to simulate a “real world” application. The comparative study highlights the similarities between the two techniques but also some differences that it is important to bear in mind when the methods are used in practical drug discovery applications.