DTox: A deep neural network-based in visio lens for large scale toxicogenomics data

With the advancement of large-scale omics technologies, particularly transcriptomics data sets on drug and treatment response repositories available in public domain, toxicogenomics has emerged as a key field in safety pharmacology and chemical risk assessment. Traditional statistics-based bioinformatics analysis poses challenges in its application across multidimensional toxicogenomic data, including administration time, dosage, and gene expression levels. Motivated by the visual inspection workflow of field experts to augment their efficiency of screening significant genes to derive meaningful insights, together with the ability of deep neural architectures to learn the image signals, we developed DTox, a deep neural network-based in visio approach. Using the Percellome toxicogenomics database, instead of utilizing the numerical gene expression values of the transcripts (gene probes of the microarray) for dose-time combinations, DTox learned the image representation of 3D surface plots of distinct time and dosage data points to train the classifier on the experts' labels of gene probe significance. DTox outperformed statistical threshold-based bioinformatics and machine learning approaches based on numerical expression values. This result shows the ability of image-driven neural networks to overcome the limitations of classical numeric value-based approaches. Further, by augmenting the model with explainability modules, our study showed the potential to reveal the visual analysis process of human experts in toxicogenomics through the model weights. While the current work demonstrates the application of the DTox model in toxicogenomic studies, it can be further generalized as an in visio approach for multi-dimensional numeric data with applications in various fields in medical data sciences.

A multi-label learning model for predicting drug-induced pathology in multi-organ based on toxicogenomics data

Drug-induced toxicity damages the health and is one of the key factors causing drug withdrawal from the market. It is of great significance to identify drug-induced target-organ toxicity, especially the detailed pathological findings, which are crucial for toxicity assessment, in the early stage of drug development process. A large variety of studies have devoted to identify drug toxicity. However, most of them are limited to single organ or only binary toxicity. Here we proposed a novel multi-label learning model named Att-RethinkNet, for predicting drug-induced pathological findings targeted on liver and kidney based on toxicogenomics data. The Att-RethinkNet is equipped with a memory structure and can effectively use the label association information. Besides, attention mechanism is embedded to focus on the important features and obtain better feature presentation. Our Att-RethinkNet is applicable in multiple organs and takes account the compound type, dose, and administration time, so it is more comprehensive and generalized. And more importantly, it predicts multiple pathological findings at the same time, instead of predicting each pathology separately as the previous model did. To demonstrate the effectiveness of the proposed model, we compared the proposed method with a series of state-of-the-arts methods. Our model shows competitive performance and can predict potential hepatotoxicity and nephrotoxicity in a more accurate and reliable way. The implementation of the proposed method is available at https://github.com/RanSuLab/Drug-Toxicity-Prediction-MultiLabel.

Genome-wide gene expression analysis reveals molecular insights into the drug-induced toxicity of nephrotoxic agents

Drug-induced nephrotoxicity is frequently reported. However, the mechanisms underlying nephrotoxic medications and their overlapping molecular events, which might have therapeutic value, are unclear. We performed a genome-wide analysis of gene expression and a gene set enrichment analysis to identify common and unique pathways associated with the toxicity of colistin, ifosfamide, indomethacin, and puromycin. Rats were randomly allocated into the treatment or control group. The treatment group received a toxic dose once daily of each investigated drug for 1 week. Differentially expressed genes were found in the drug-treated kidney and liver compared to the control, except for colistin in the liver. Upregulated pathways were mainly related to cell death, cell cycle, protein synthesis, and immune response modulation in the kidney. Cell cycle was upregulated by all drugs. Downregulated pathways were associated with carbon metabolism, amino acid metabolism, and fatty acid metabolism. Indomethacin, colistin, and puromycin shared the most altered pathways in the kidney. Ifosfamide and indomethacin affected molecular processes greatly in the liver. Our findings provide insight into the mechanisms underlying the renal and hepatic adverse effects of the four drugs. Further investigation should explore the combinatory drug therapies that attenuate the toxic effects and maximize the effectiveness of nephrotoxic drugs.

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.

Assessment of Drugs Toxicity and Associated Biomarker Genes Using Hierarchical Clustering

Background and objectives: Assessment of drugs toxicity and associated biomarker genes is one of the most important tasks in the pre-clinical phase of drug development pipeline as well as in toxicogenomic studies. There are few statistical methods for the assessment of doses of drugs (DDs) toxicity and their associated biomarker genes. However, these methods consume more time for computation of the model parameters using the EM (expectation-maximization) based iterative approaches. To overcome this problem, in this paper, an attempt is made to propose an alternative approach based on hierarchical clustering (HC) for the same purpose. Methods and materials: There are several types of HC approaches whose performance depends on different similarity/distance measures. Therefore, we explored suitable combinations of distance measures and HC methods based on Japanese Toxicogenomics Project (TGP) datasets for better clustering/co-clustering between DDs and genes as well as to detect toxic DDs and their associated biomarker genes. Results: We observed that Word’s HC method with each of Euclidean, Manhattan, and Minkowski distance measures produces better clustering/co-clustering results. For an example, in the case of the glutathione metabolism pathway (GMP) dataset LOC100359539/Rrm2, Gpx6, RGD1562107, Gstm4, Gstm3, G6pd, Gsta5, Gclc, Mgst2, Gsr, Gpx2, Gclm, Gstp1, LOC100912604/Srm, Gstm4, Odc1, Gsr, Gss are the biomarker genes and Acetaminophen_Middle, Acetaminophen_High, Methapyrilene_High, Nitrofurazone_High, Nitrofurazone_Middle, Isoniazid_Middle, Isoniazid_High are their regulatory (associated) DDs explored by our proposed co-clustering algorithm based on the distance and HC method combination Euclidean: Word. Similarly, for the peroxisome proliferator-activated receptor signaling pathway (PPAR-SP) dataset Cpt1a, Cyp8b1, Cyp4a3, Ehhadh, Plin5, Plin2, Fabp3, Me1, Fabp5, LOC100910385, Cpt2, Acaa1a, Cyp4a1, LOC100365047, Cpt1a, LOC100365047, Angptl4, Aqp7, Cpt1c, Cpt1b, Me1 are the biomarker genes and Aspirin_Low, Aspirin_Middle, Aspirin_High, Benzbromarone_Middle, Benzbromarone_High, Clofibrate_Middle, Clofibrate_High, WY14643_Low, WY14643_High, WY14643_Middle, Gemfibrozil_Middle, Gemfibrozil_High are their regulatory DDs. Conclusions: Overall, the methods proposed in this article, co-cluster the genes and DDs as well as detect biomarker genes and their regulatory DDs simultaneously consuming less time compared to other mentioned methods. The results produced by the proposed methods have been validated by the available literature and functional annotation.

Developing a Multi-Dose Computational Model for Drug-Induced Hepatotoxicity Prediction Based on Toxicogenomics Data

Drug-induced hepatotoxicity may cause acute and chronic liver disease, leading to great concern for patient safety. It is also one of the main reasons for drug withdrawal from the market. Toxicogenomics data has been widely used in hepatotoxicity prediction. In our study, we proposed a multi-dose computational model to predict the drug-induced hepatotoxicity based on gene expression and toxicity data. The dose/concentration information after drug treatment is fully utilized in our study based on the dose-response curve, thus a more informative representative of the dose-response relationship is considered. We also proposed a new feature selection method, named MEMO, which is also one important aspect of our multi-dose model in our study, to deal with the high-dimensional toxicogenomics data. We validated the proposed model using the TG-GATEs, which is a large database recording toxicogenomics data from multiple views. The experimental results show that the drug-induced hepatotoxicity can be predicted with high accuracy and efficiency using the proposed predictive model.

In Silico Toxicology Data Resources to Support Read-Across and (Q)SAR

A plethora of databases exist online that can assist in in silico chemical or drug safety assessment. However, a systematic review and grouping of databases, based on purpose and information content, consolidated in a single source, has been lacking. To resolve this issue, this review provides a comprehensive listing of the key in silico data resources relevant to: chemical identity and properties, drug action, toxicology (including nano-material toxicity), exposure, omics, pathways, Absorption, Distribution, Metabolism and Elimination (ADME) properties, clinical trials, pharmacovigilance, patents-related databases, biological (genes, enzymes, proteins, other macromolecules etc.) databases, protein-protein interactions (PPIs), environmental exposure related, and finally databases relating to animal alternatives in support of 3Rs policies. More than nine hundred databases were identified and reviewed against criteria relating to accessibility, data coverage, interoperability or application programming interface (API), appropriate identifiers, types of in vitro, in vivo,-clinical or other data recorded and suitability for modelling, read-across, or similarity searching. This review also specifically addresses the need for solutions for mapping and integration of databases into a common platform for better translatability of preclinical data to clinical data.

MinE-RFE: determine the optimal subset from RFE by minimizing the subset-accuracy–defined energy

Recursive feature elimination (RFE), as one of the most popular feature selection algorithms, has been extensively applied to bioinformatics. During the training, a group of candidate subsets are generated by iteratively eliminating the least important features from the original features. However, how to determine the optimal subset from them still remains ambiguous. Among most current studies, either overall accuracy or subset size (SS) is used to select the most predictive features. Using which one or both and how they affect the prediction performance are still open questions. In this study, we proposed MinE-RFE, a novel RFE-based feature selection approach by sufficiently considering the effect of both factors. Subset decision problem was reflected into subset-accuracy space and became an energy-minimization problem. We also provided a mathematical description of the relationship between the overall accuracy and SS using Gaussian Mixture Models together with spline fitting. Besides, we comprehensively reviewed a variety of state-of-the-art applications in bioinformatics using RFE. We compared their approaches of deciding the final subset from all the candidate subsets with MinE-RFE on diverse bioinformatics data sets. Additionally, we also compared MinE-RFE with some well-used feature selection algorithms. The comparative results demonstrate that the proposed approach exhibits the best performance among all the approaches. To facilitate the use of MinE-RFE, we further established a user-friendly web server with the implementation of the proposed approach, which is accessible at http://qgking.wicp.net/MinE/. We expect this web server will be a useful tool for research community.

Logic programming to infer complex RNA expression patterns from RNA-seq data

To meet the increasing demand in the field, numerous long noncoding RNA (lncRNA) databases are available. Given many lncRNAs are specifically expressed in certain cell types and/or time-dependent manners, most lncRNA databases fall short of providing such profiles. We developed a strategy using logic programming to handle the complex organization of organs, their tissues and cell types as well as gender and developmental time points. To showcase this strategy, we introduce 'RenalDB' (http://renaldb.uni-frankfurt.de), a database providing expression profiles of RNAs in major organs focusing on kidney tissues and cells. RenalDB uses logic programming to describe complex anatomy, sample metadata and logical relationships defining expression, enrichment or specificity. We validated the content of RenalDB with biological experiments and functionally characterized two long intergenic noncoding RNAs: LOC440173 is important for cell growth or cell survival, whereas PAXIP1-AS1 is a regulator of cell death. We anticipate RenalDB will be used as a first step toward functional studies of lncRNAs in the kidney.

Robust identification of significant interactions between toxicogenomic biomarkers and their regulatory chemical compounds using logistic moving range chart

Identification of significant interactions between genes and chemical compounds/drugs is an important issue in toxicogenomic studies as well as in drug discovery and development. There are some online and offline computational tools for toxicogenomic data analysis to identify the biomarker genes and their regulatory chemical compounds/drugs. However, none of the researchers has considered yet the identification of significant interactions between genes and compounds. Therefore, in this paper, we have discussed two approaches namely moving range chart (MRC) and logistic moving range chart (LMRC) for the identification of significant up-regulatory (UpR) and down-regulatory (DnR) gene-compound interactions as well as toxicogenomic biomarkers and their regulatory chemical compounds/drugs. We have investigated the performance of both MRC and LMRC approaches using simulated datasets. Simulation results show that both approaches perform almost equally in absence of outliers. However, in presence of outliers, the LMRC shows much better performance than the MRC. In case of real life toxicogenomic data analysis, the proposed LMRC approach detected some important down-regulated biomarker genes those were not detected by other approaches. Therefore, in this paper, our proposal is to use LMRC for robust identification of significant interactions between genes and chemical compounds/drugs.

Robust Co-clustering to Discover Toxicogenomic Biomarkers and Their Regulatory Doses of Chemical Compounds Using Logistic Probabilistic Hidden Variable Model

Detection of biomarker genes and their regulatory doses of chemical compounds (DCCs) is one of the most important tasks in toxicogenomic studies as well as in drug design and development. There is an online computational platform "Toxygates" to identify biomarker genes and their regulatory DCCs by co-clustering approach. Nevertheless, the algorithm of that platform based on hierarchical clustering (HC) does not share gene-DCC two-way information simultaneously during co-clustering between genes and DCCs. Also it is sensitive to outlying observations. Thus, this platform may produce misleading results in some cases. The probabilistic hidden variable model (PHVM) is a more effective co-clustering approach that share two-way information simultaneously, but it is also sensitive to outlying observations. Therefore, in this paper we have proposed logistic probabilistic hidden variable model (LPHVM) for robust co-clustering between genes and DCCs, since gene expression data are often contaminated by outlying observations. We have investigated the performance of the proposed LPHVM co-clustering approach in a comparison with the conventional PHVM and Toxygates co-clustering approaches using simulated and real life TGP gene expression datasets, respectively. Simulation results show that the proposed method improved the performance over the conventional PHVM in presence of outliers; otherwise, it keeps equal performance. In the case of real life TGP data analysis, three DCCs (glibenclamide-low, perhexilline-low, and hexachlorobenzene-medium) for glutathione metabolism pathway dataset as well as two DCCs (acetaminophen-medium and methapyrilene-low) for PPAR signaling pathway dataset were incorrectly co-clustered by the Toxygates online platform, while only one DCC (hexachlorobenzene-low) for glutathione metabolism pathway was incorrectly co-clustered by the proposed LPHVM approach. Our findings from the real data analysis are also supported by the other findings in the literature.

Gene expression profiling toward the next generation safety control of influenza vaccines and adjuvants in Japan

Influenza becomes epidemic worldwide every year, and many individuals receive vaccination annually. Quality control relating to safety and potency of influenza vaccines is important to maintain public confidence. The safety of influenza vaccines has been assessed by clinical trials, and animal safety tests are performed to monitor the consistent quality between vaccines used for clinical trials and marketing; the biological responses in vaccinated animals are evaluated, including changes in body weight and white blood cell count. Animal safety tests have been contributing to the quality relating to the safety of influenza vaccines for decades, but improvements are needed. Although precise mechanisms involving biological changes in animal safety tests have not been fully elucidated, the application of cDNA microarray technology make it possible to reliably identify genes related to biological responses in vaccinated animals. From analysis of the expression profile of >10,000 genes of lung in animals treated with an inactivated whole virion influenza vaccine, we identified 17 marker genes whose expression patterns correlated well to changes in body weight and leukocyte count in vaccinated animals. In influenza HA vaccine-treated animals exhibiting subtle changes in biological responses, a robust expression pattern of marker genes was found. Furthermore, these marker genes could also be used in the evaluation of adjuvanted influenza vaccines. The expression profile of marker genes is expected to be an alternative indicator for safety control of various influenza vaccines conferring high sensitivity and short turnaround time. Thus, gene expression profiling may be a powerful tool for safety control of vaccines in the future.

Toxic Dose prediction of Chemical Compounds to Biomarkers using an ANOVA based Gene Expression Analysis

The aim of toxicogenomic studies is to optimize the toxic dose levels of chemical compounds (CCs) and their regulated biomarker genes. This is also crucial in drug discovery and development. There are popular online computational tools such as ToxDB and Toxygates to identify toxicogenomic biomarkers using t-test. However, they are not suitable for the identification of biomarker gene regulatory dose of corresponding CCs. Hence, we describe a one-way ANOVA model together with Tukey's HSD test for the identification of toxicogenomic biomarker genes and their influencing CC dose with improved efficiency. Glutathione metabolism pathway data analysis shows high and middle dose for acetaminophen, and nitrofurazone as well as high dose for methapyrilene as significant toxic CC dose. The corresponding regulated top seven toxicogenomic biomarker genes found in this analysis is Gstp1, Gsr, Mgst2, Gclm, G6pd, Gsta5 and Gclc.

Decision Variants for the Automatic Determination of Optimal Feature Subset in RF-RFE

Feature selection, which identifies a set of most informative features from the original feature space, has been widely used to simplify the predictor. Recursive feature elimination (RFE), as one of the most popular feature selection approaches, is effective in data dimension reduction and efficiency increase. A ranking of features, as well as candidate subsets with the corresponding accuracy, is produced through RFE. The subset with highest accuracy (HA) or a preset number of features (PreNum) are often used as the final subset. However, this may lead to a large number of features being selected, or if there is no prior knowledge about this preset number, it is often ambiguous and subjective regarding final subset selection. A proper decision variant is in high demand to automatically determine the optimal subset. In this study, we conduct pioneering work to explore the decision variant after obtaining a list of candidate subsets from RFE. We provide a detailed analysis and comparison of several decision variants to automatically select the optimal feature subset. Random forest (RF)-recursive feature elimination (RF-RFE) algorithm and a voting strategy are introduced. We validated the variants on two totally different molecular biology datasets, one for a toxicogenomic study and the other one for protein sequence analysis. The study provides an automated way to determine the optimal feature subset when using RF-RFE.

Interactive Toxicogenomics: Gene set discovery, clustering and analysis in Toxygates

Toxygates was originally released as a user-friendly interface to enhance the accessibility of the large-scale toxicogenomics database, Open TG-GATEs, generated by the Japanese Toxicogenomics Project. Since the original release, significant new functionality has been added to enable users to perform sophisticated computational analysis with only modest bioinformatics skills. The new features include an orthologous mode for data comparison among different species, interactive clustering and heatmap visualisation, enrichment analysis of gene sets, and user data uploading. In a case study, we use these new functions to study the hepatotoxicity of peroxisome proliferator-activated receptor alpha (PPARα) agonist WY-14643. Our findings suggest that WY-14643 caused hypertrophy in the bile duct by intracellular Ca²⁺ dysregulation, which resulted in the induction of genes in a non-canonical WNT/Ca²⁺ signalling pathway. With this new release of Toxygates, we provide a suite of tools that allow anyone to carry out in-depth analysis of toxicogenomics in Open TG-GATEs, and of any other dataset that is uploaded.

ZBIT Bioinformatics Toolbox: A Web-Platform for Systems Biology and Expression Data Analysis

Bioinformatics analysis has become an integral part of research in biology. However, installation and use of scientific software can be difficult and often requires technical expert knowledge. Reasons are dependencies on certain operating systems or required third-party libraries, missing graphical user interfaces and documentation, or nonstandard input and output formats. In order to make bioinformatics software easily accessible to researchers, we here present a web-based platform. The Center for Bioinformatics Tuebingen (ZBIT) Bioinformatics Toolbox provides web-based access to a collection of bioinformatics tools developed for systems biology, protein sequence annotation, and expression data analysis. Currently, the collection encompasses software for conversion and processing of community standards SBML and BioPAX, transcription factor analysis, and analysis of microarray data from transcriptomics and proteomics studies. All tools are hosted on a customized Galaxy instance and run on a dedicated computation cluster. Users only need a web browser and an active internet connection in order to benefit from this service. The web platform is designed to facilitate the usage of the bioinformatics tools for researchers without advanced technical background. Users can combine tools for complex analyses or use predefined, customizable workflows. All results are stored persistently and reproducible. For each tool, we provide documentation, tutorials, and example data to maximize usability. The ZBIT Bioinformatics Toolbox is freely available at https://webservices.cs.uni-tuebingen.de/.

Predicting target proteins for drug candidate compounds based on drug-induced gene expression data in a chemical structure-independent manner

BACKGROUND

Phenotype-based high-throughput screening is a useful technique for identifying drug candidate compounds that have a desired phenotype. However, the molecular mechanisms of the hit compounds remain unknown, and substantial effort is required to identify the target proteins associated with the phenotype.

METHODS

In this study, we propose a new method to predict target proteins of drug candidate compounds based on drug-induced gene expression data in Connectivity Map and a machine learning classification technique, which we call the "transcriptomic approach."

RESULTS

Unlike existing methods such as the chemogenomic approach, the transcriptomic approach enabled the prediction of target proteins without dependence on prior knowledge of compound chemical structures. The prediction accuracy of the chemogenomic approach was highly depended on compounds structure similarities in data sets. In contrast, the prediction accuracy of the transcriptomic approach was maintained at a sufficient level, even for benchmark data consisting of structurally diverse compounds.

CONCLUSIONS

The transcriptomic approach reported here is expected to be a useful tool for structure-independent prediction of target proteins for drug candidate compounds.

Workshop report: Identifying opportunities for global integration of toxicogenomics databases, 26-27 June 2013, Research Triangle Park, NC, USA

A joint US-EU workshop on enhancing data sharing and exchange in toxicogenomics was held at the National Institute for Environmental Health Sciences. Currently, efficient reuse of data is hampered by problems related to public data availability, data quality, database interoperability (the ability to exchange information), standardization and sustainability. At the workshop, experts from universities and research institutes presented databases, studies, organizations and tools that attempt to deal with these problems. Furthermore, a case study showing that combining toxicogenomics data from multiple resources leads to more accurate predictions in risk assessment was presented. All participants agreed that there is a need for a web portal describing the diverse, heterogeneous data resources relevant for toxicogenomics research. Furthermore, there was agreement that linking more data resources would improve toxicogenomics data analysis. To outline a roadmap to enhance interoperability between data resources, the participants recommend collecting user stories from the toxicogenomics research community on barriers in data sharing and exchange currently hampering answering to certain research questions. These user stories may guide the prioritization of steps to be taken for enhancing integration of toxicogenomics databases.

Open TG-GATEs: a large-scale toxicogenomics database

Toxicogenomics focuses on assessing the safety of compounds using gene expression profiles. Gene expression signatures from large toxicogenomics databases are expected to perform better than small databases in identifying biomarkers for the prediction and evaluation of drug safety based on a compound's toxicological mechanisms in animal target organs. Over the past 10 years, the Japanese Toxicogenomics Project consortium (TGP) has been developing a large-scale toxicogenomics database consisting of data from 170 compounds (mostly drugs) with the aim of improving and enhancing drug safety assessment. Most of the data generated by the project (e.g. gene expression, pathology, lot number) are freely available to the public via Open TG-GATEs (Toxicogenomics Project-Genomics Assisted Toxicity Evaluation System). Here, we provide a comprehensive overview of the database, including both gene expression data and metadata, with a description of experimental conditions and procedures used to generate the database. Open TG-GATEs is available from https://toxico.nibiohn.go.jp/english/index.html.

Cancer biology, toxicology and alternative methods development go hand-in-hand

Toxicological research faces the challenge of integrating knowledge from diverse fields and novel technological developments generally in the biological and medical sciences. We discuss herein the fact that the multiple facets of cancer research, including discovery related to mechanisms, treatment and diagnosis, overlap many up and coming interest areas in toxicology, including the need for improved methods and analysis tools. Common to both disciplines, in vitro and in silico methods serve as alternative investigation routes to animal studies. Knowledge on cancer development helps in understanding the relevance of chemical toxicity studies in cell models, and many bioinformatics-based cancer biomarker discovery tools are also applicable to computational toxicology. Robotics-aided, cell-based, high-throughput screening, microscale immunostaining techniques and gene expression profiling analyses are common tools in cancer research, and when sequentially combined, form a tiered approach to structured safety evaluation of thousands of environmental agents, novel chemicals or engineered nanomaterials. Comprehensive tumour data collections in databases have been translated into clinically useful data, and this concept serves as template for computer-driven evaluation of toxicity data into meaningful results. Future 'cancer research-inspired knowledge management' of toxicological data will aid the translation of basic discovery results and chemicals- and materials-testing data to information relevant to human health and environmental safety.