Developing a GNN-based AI model to predict mitochondrial toxicity using the bagging method.

Mitochondrial toxicity has been implicated in the development of various toxicities, including hepatotoxicity. Therefore, mitochondrial toxicity has become a major screening factor in the early discovery phase of drug development. Several models have been developed to predict mitochondrial toxicity based on chemical structures. However, they only provide a binary classification of positive or negative results and do not provide the substructures that contribute to a positive decision. Therefore, we developed an artificial intelligence (AI) model to predict mitochondrial toxicity and visualize structural alerts. To construct the model, we used the open-source software library kMoL, which employs a graph neural network approach that allows learning from chemical structure data. We also utilized the integrated gradient method, which enables the visualization of substructures that contribute to positive results. The dataset used to construct the AI model exhibited a significant imbalance, with significantly more negative than positive data. To address this, we employed the bagging method, which resulted in a model with high predictive performance, as evidenced by an F1 score of 0.839. This model can also be used to visualize substructures that contribute to mitochondrial toxicity using the integrated gradient method. Our AI model predicts mitochondrial toxicity based on chemical structures and may contribute to screening mitochondrial toxicity in the early stages of drug discovery.

Development of a GCN-based model to predict in vitro phototoxicity from the chemical structure and HOMO-LUMO gap.

The interaction between sunlight and drugs can lead to phototoxicity in patients who have received such drugs. Phototoxicity assessment is a regulatory requirement globally and one of the main toxicity screening steps in the early stages of drug discovery. An in silico-in vitro approach has been utilized mainly for toxicology assessments at these stages. Although several quantitative structure-activity relationship (QSAR) models for phototoxicity have been developed, in silico technology to evaluate phototoxicity has not been well established. In this study, we attempted to develop an artificial intelligence (AI) model to predict the in vitro Neutral Red Uptake Phototoxicity Test results from a chemical structure and its derived information. To accomplish this, we utilized an open-source software library, kMoL. kMoL employs a graph convolutional neural networks (GCN) approach, which allows it to learn the data for the specified chemical structure. kMoL also utilizes the integrated gradient (IG) method, enabling it to visually display the substructures contributing to any positive results. To construct this AI model, we used only the chemical structure as a basis, then added the descriptors and the HOMO-LUMO gap, which was obtained from quantum chemical calculations. As a result, the assortment of chemical structures and the HOMO-LUMO gap produced an AI model with high discrimination performance, and an F1 score of 0.857. Additionally, our AI model could visualize the substructures involved in phototoxicity using the IG method. Our AI model can be applied as a toxicity screening method and could enhance productivity in drug development.

Quantifying the relative immune cell activation from whole tissue/organ-derived differentially expressed gene data.

Evaluation of immune responses in individual immune cell types is important for the development of new medicines. Here, we propose a computational method designated ICEPOP (Immune CEll POPulation) to estimate individual immune cell type responses from bulk tissue and organ samples. The relative gene responses are scored for each cell type by using the data from differentially expressed genes derived from control- vs drug-treated sample pairs, and the data from public databases including ImmGen and IRIS, which contain gene expression profiles of a variety of immune cells. By ICEPOP, we analysed cell responses induced by vaccine-adjuvants in the mouse spleen, and extended the analyses to human peripheral blood mononuclear cells and gut biopsy samples focusing on human papilloma virus vaccination and inflammatory bowel disease treatment with Infliximab. In both mouse and human datasets, our method reliably quantified the responding immune cell types and provided insightful information, demonstrating that our method is useful to evaluate immune responses from bulk sample-derived gene expression data. ICEPOP is available as an interactive web site ( https://vdynamics.shinyapps.io/icepop/ ) and Python package ( https://github.com/ewijaya/icepop ).

Isoflurane is a suitable alternative to ether for anesthetizing rats prior to euthanasia for gene expression analysis.

Diethyl ether (ether) had been widely used in Japan for anesthesia, despite its explosive properties and toxicity to both humans and animals. We also had used ether as an anesthetic for euthanizing rats for research in the Toxicogenomics Project (TGP). Because the use of ether for these purposes will likely cease, it is required to select an alternative anesthetic which is validated for consistency with existing TGP data acquired under ether anesthesia. We therefore compared two alternative anesthetic candidates, isoflurane and pentobarbital, with ether in terms of hematological findings, serum biochemical parameters, and gene expressions. As a result, few differences among the three agents were observed. In hematological and serum biochemistry analysis, no significant changes were found. In gene expression analysis, four known genes were extracted as differentially expressed genes in the liver of rats anesthetized with ether, isoflurane, or pentobarbital. However, no significant relationships were detected using gene ontology, pathway, or gene enrichment analyses by DAVID and TargetMine. Surprisingly, although it was expected that the lung would be affected by administration via inhalation, only one differentially expressed gene was extracted in the lung. Taken together, our data indicate that there are no significant differences among ether, isoflurane, and pentobarbital with respect to effects on hematological parameters, serum biochemistry parameters, and gene expression. Based on its smallest affect to existing data and its safety profile for humans and animals, we suggest isoflurane as a suitable alternative anesthetic for use in rat euthanasia in toxicogenomics analysis.

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.

Advax, a Delta Inulin Microparticle, Potentiates In-built Adjuvant Property of Co-administered Vaccines.

Advax, a delta inulin-derived microparticle, has been developed as an adjuvant for several vaccines. However, its immunological characteristics and potential mechanism of action are yet to be elucidated. Here, we show that Advax behaves as a type-2 adjuvant when combined with influenza split vaccine, a T helper (Th)2-type antigen, but behaves as a type-1 adjuvant when combined with influenza inactivated whole virion (WV), a Th1-type antigen. In addition, an adjuvant effect was not observed when Advax-adjuvanted WV vaccine was used to immunize toll-like receptor (TLR) 7 knockout mice which are unable to respond to RNA contained in WV antigen. Similarly, no adjuvant effect was seen when Advax was combined with endotoxin-free ovalbumin, a neutral Th0-type antigen. An adjuvant effect was also not seen in tumor necrosis factor (TNF)-α knockout mice, and the adjuvant effect required the presences of dendritic cells (DCs) and phagocytic macrophages. Therefore, unlike other adjuvants, Advax potentiates the intrinsic or in-built adjuvant property of co-administered antigens. Hence, Advax is a unique class of adjuvant which can potentiate the intrinsic adjuvant feature of the vaccine antigens through a yet to be determined mechanism.

Molecular Markers of Tubulointerstitial Fibrosis and Tubular Cell Damage in Patients with Chronic Kidney Disease.

In chronic kidney disease (CKD), progressive nephron loss causes glomerular sclerosis, as well as tubulointerstitial fibrosis and progressive tubular injury. In this study, we aimed to identify molecular changes that reflected the histopathological progression of renal tubulointerstitial fibrosis and tubular cell damage. A discovery set of renal biopsies were obtained from 48 patients with histopathologically confirmed CKD, and gene expression profiles were determined by microarray analysis. The results indicated that hepatitis A virus cellular receptor 1 (also known as Kidney Injury Molecule-1, KIM-1), lipocalin 2 (also known as neutrophil gelatinase-associated lipocalin, NGAL), SRY-box 9, WAP four-disulfide core domain 2, and NK6 homeobox 2 were differentially expressed in CKD. Their expression levels correlated with the extent of tubulointerstitial fibrosis and tubular cell injury, determined by histopathological examination. The expression of these 5 genes was also increased as kidney damage progressed in a rodent unilateral ureteral obstruction model of CKD. We calculated a molecular score using the microarray gene expression profiles of the biopsy specimens. The composite area under the receiver operating characteristics curve plotted using this molecular score showed a high accuracy for diagnosing tubulointerstitial fibrosis and tubular cell damage. The robust sensitivity of this score was confirmed in a validation set of 5 individuals with CKD. These findings identified novel molecular markers with the potential to contribute to the detection of tubular cell damage and tubulointerstitial fibrosis in the kidney.

Functional characterization of importin α8 as a classical nuclear localization signal receptor.

Importin α8 has recently been identified as an importin α family member based on its primary structure and binding ability to importin ß1 and to several karyophilic proteins. However, there has been no experimental evidence that importin α8 actually functions in the nuclear transport of classical nuclear localization signal (cNLS)-containing cargo. Here, using an in vitro transport assay, we demonstrate that purified recombinant importin α8 can transport SV40T antigen cNLS-containing cargo into the nucleus of HeLa cells, in conjunction with importin ß1. Pull-down assays, followed by mass spectrometry analysis, identified 179 putative importin α8-binding proteins, only 62 of which overlap with those of importin α1, the closest importin α family member. Among the importin α8-binding candidates, we showed that DNA damage-binding protein 2 (DDB2) was actually transported into the nucleus via the importin α8/ß1 pathway. Furthermore, we found that the other subtypes of importin α, which were also identified as importin α8-binding candidates, indeed form heterodimers with importin α8. Notably, we found that these importin α8-containing heterodimers were more stable in the presence of cNLS-substrates than heterodimers containing importin α1. From these findings, we propose that importin α8 functions as a cNLS receptor with distinct cargo specificity, and that heterodimerization by importin α8 is a novel regulatory mode of cNLS binding, in addition to the autoinhibitory regulation by the importin ß binding domain.

Implementation of linked data in the life sciences at BioHackathon 2011.

BACKGROUND: Linked Data has gained some attention recently in the life sciences as an effective way to provide and share data. As a part of the Semantic Web, data are linked so that a person or machine can explore the web of data. Resource Description Framework (RDF) is the standard means of implementing Linked Data. In the process of generating RDF data, not only are data simply linked to one another, the links themselves are characterized by ontologies, thereby allowing the types of links to be distinguished. Although there is a high labor cost to define an ontology for data providers, the merit lies in the higher level of interoperability with data analysis and visualization software. This increase in interoperability facilitates the multi-faceted retrieval of data, and the appropriate data can be quickly extracted and visualized. Such retrieval is usually performed using the SPARQL (SPARQL Protocol and RDF Query Language) query language, which is used to query RDF data stores. For the database provider, such interoperability will surely lead to an increase in the number of users. RESULTS: This manuscript describes the experiences and discussions shared among participants of the week-long BioHackathon 2011 who went through the development of RDF representations of their own data and developed specific RDF and SPARQL use cases. Advice regarding considerations to take when developing RDF representations of their data are provided for bioinformaticians considering making data available and interoperable. CONCLUSIONS: Participants of the BioHackathon 2011 were able to produce RDF representations of their data and gain a better understanding of the requirements for producing such data in a period of just five days. We summarize the work accomplished with the hope that it will be useful for researchers involved in developing laboratory databases or data analysis, and those who are considering such technologies as RDF and Linked Data.

Hydroxypropyl-ß-cyclodextrin spikes local inflammation that induces Th2 cell and T follicular helper cell responses to the coadministered antigen.

Cyclodextrins are commonly used as a safe excipient to enhance the solubility and bioavailability of hydrophobic pharmaceutical agents. Their efficacies and mechanisms as drug-delivery systems have been investigated for decades, but their immunological properties have not been examined. In this study, we reprofiled hydroxypropyl-ß-cyclodextrin (HP-ß-CD) as a vaccine adjuvant and found that it acts as a potent and unique adjuvant. HP-ß-CD triggered the innate immune response at the injection site, was trapped by MARCO(+) macrophages, increased Ag uptake by dendritic cells, and facilitated the generation of T follicular helper cells in the draining lymph nodes. It significantly enhanced Ag-specific Th2 and IgG Ab responses as potently as did the conventional adjuvant, aluminum salt (alum), whereas its ability to induce Ag-specific IgE was less than that of alum. At the injection site, HP-ß-CD induced the temporary release of host dsDNA, a damage-associated molecular pattern. DNase-treated mice, MyD88-deficient mice, and TBK1-deficient mice showed significantly reduced Ab responses after immunization with this adjuvant. Finally, we demonstrated that HP-ß-CD-adjuvanted influenza hemagglutinin split vaccine protected against a lethal challenge with a clinically isolated pandemic H1N1 influenza virus, and the adjuvant effect of HP-ß-CD was demonstrated in cynomolgus macaques. Our results suggest that HP-ß-CD acts as a potent MyD88- and TBK1-dependent T follicular helper cell adjuvant and is readily applicable to various vaccines.

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 http://toxico.nibio.go.jp/english/index.html.

Basis for substrate recognition and distinction by matrix metalloproteinases.

Genomic sequencing and structural genomics produced a vast amount of sequence and structural data, creating an opportunity for structure-function analysis in silico [Radivojac P, et al. (2013) Nat Methods 10(3):221-227]. Unfortunately, only a few large experimental datasets exist to serve as benchmarks for function-related predictions. Furthermore, currently there are no reliable means to predict the extent of functional similarity among proteins. Here, we quantify structure-function relationships among three phylogenetic branches of the matrix metalloproteinase (MMP) family by comparing their cleavage efficiencies toward an extended set of phage peptide substrates that were selected from ∼ 64 million peptide sequences (i.e., a large unbiased representation of substrate space). The observed second-order rate constants [k(obs)] across the substrate space provide a distance measure of functional similarity among the MMPs. These functional distances directly correlate with MMP phylogenetic distance. There is also a remarkable and near-perfect correlation between the MMP substrate preference and sequence identity of 50-57 discontinuous residues surrounding the catalytic groove. We conclude that these residues represent the specificity-determining positions (SDPs) that allowed for the expansion of MMP proteolytic function during evolution. A transmutation of only a few selected SDPs proximal to the bound substrate peptide, and contributing the most to selectivity among the MMPs, is sufficient to enact a global change in the substrate preference of one MMP to that of another, indicating the potential for the rational and focused redesign of cleavage specificity in MMPs.

BioHackathon series in 2011 and 2012: penetration of ontology and linked data in life science domains.

The application of semantic technologies to the integration of biological data and the interoperability of bioinformatics analysis and visualization tools has been the common theme of a series of annual BioHackathons hosted in Japan for the past five years. Here we provide a review of the activities and outcomes from the BioHackathons held in 2011 in Kyoto and 2012 in Toyama. In order to efficiently implement semantic technologies in the life sciences, participants formed various sub-groups and worked on the following topics: Resource Description Framework (RDF) models for specific domains, text mining of the literature, ontology development, essential metadata for biological databases, platforms to enable efficient Semantic Web technology development and interoperability, and the development of applications for Semantic Web data. In this review, we briefly introduce the themes covered by these sub-groups. The observations made, conclusions drawn, and software development projects that emerged from these activities are discussed.

Toxygates: interactive toxicity analysis on a hybrid microarray and linked data platform.

MOTIVATION: In early stage drug development, it is desirable to assess the toxicity of compounds as quickly as possible. Biomarker genes can help predict whether a candidate drug will adversely affect a given individual, but they are often difficult to discover. In addition, the mechanism of toxicity of many drugs and common compounds is not yet well understood. The Japanese Toxicogenomics Project provides a large database of systematically collected microarray samples from rats (liver, kidney and primary hepatocytes) and human cells (primary hepatocytes) after exposure to 170 different compounds in different dosages and at different time intervals. However, until now, no intuitive user interface has been publically available, making it time consuming and difficult for individual researchers to explore the data. RESULTS: We present Toxygates, a user-friendly integrated analysis platform for this database. Toxygates combines a large microarray dataset with the ability to fetch semantic linked data, such as pathways, compound-protein interactions and orthologs, on demand. It can also perform pattern-based compound ranking with respect to the expression values of a set of relevant candidate genes. By using Toxygates, users can freely interrogate the transcriptome's response to particular compounds and conditions, which enables deep exploration of toxicity mechanisms.

Sagace: a web-based search engine for biomedical databases in Japan.

BACKGROUND: In the big data era, biomedical research continues to generate a large amount of data, and the generated information is often stored in a database and made publicly available. Although combining data from multiple databases should accelerate further studies, the current number of life sciences databases is too large to grasp features and contents of each database. FINDINGS: We have developed Sagace, a web-based search engine that enables users to retrieve information from a range of biological databases (such as gene expression profiles and proteomics data) and biological resource banks (such as mouse models of disease and cell lines). With Sagace, users can search more than 300 databases in Japan. Sagace offers features tailored to biomedical research, including manually tuned ranking, a faceted navigation to refine search results, and rich snippets constructed with retrieved metadata for each database entry. CONCLUSIONS: Sagace will be valuable for experts who are involved in biomedical research and drug development in both academia and industry. Sagace is freely available at http://sagace.nibio.go.jp/en/.

Evaluation of DNA microarray results in the Toxicogenomics Project (TGP) consortium in Japan.

An important technology used in toxicogenomic drug discovery research is the microarray, which enables researchers to simultaneously analyze the expression of a large number of genes. To build a database and data analysis system for use in assessing the safety of drugs and drug candidates, in 2002 we conducted a 5-year collaborative study in the Toxicogenomics Project (TGP1) in Japan. Experimental data generated by such studies must be validated by different laboratories for robust and accurate analysis. For this purpose, we conducted intra- and inter-laboratory validation studies with participating companies in the second collaborative study in the Toxicogenomics Project (TGP2). Gene expression in the liver of rats treated with acetaminophen (APAP) was independently examined by the participating companies using Affymetrix GeneChip microarrays. The intra- and inter-laboratory reproducibility of the data was evaluated using hierarchical clustering analysis. The toxicogenomics results were highly reproducible, indicating that the gene expression data generated in our TGP1 project is reliable and compatible with the data generated by the participating laboratories.

Prediction model of potential hepatocarcinogenicity of rat hepatocarcinogens using a large-scale toxicogenomics database.

The present study was performed to develop a robust gene-based prediction model for early assessment of potential hepatocarcinogenicity of chemicals in rats by using our toxicogenomics database, TG-GATEs (Genomics-Assisted Toxicity Evaluation System developed by the Toxicogenomics Project in Japan). The positive training set consisted of high- or middle-dose groups that received 6 different non-genotoxic hepatocarcinogens during a 28-day period. The negative training set consisted of high- or middle-dose groups of 54 non-carcinogens. Support vector machine combined with wrapper-type gene selection algorithms was used for modeling. Consequently, our best classifier yielded prediction accuracies for hepatocarcinogenicity of 99% sensitivity and 97% specificity in the training data set, and false positive prediction was almost completely eliminated. Pathway analysis of feature genes revealed that the mitogen-activated protein kinase p38- and phosphatidylinositol-3-kinase-centered interactome and the v-myc myelocytomatosis viral oncogene homolog-centered interactome were the 2 most significant networks. The usefulness and robustness of our predictor were further confirmed in an independent validation data set obtained from the public database. Interestingly, similar positive predictions were obtained in several genotoxic hepatocarcinogens as well as non-genotoxic hepatocarcinogens. These results indicate that the expression profiles of our newly selected candidate biomarker genes might be common characteristics in the early stage of carcinogenesis for both genotoxic and non-genotoxic carcinogens in the rat liver. Our toxicogenomic model might be useful for the prospective screening of hepatocarcinogenicity of compounds and prioritization of compounds for carcinogenicity testing.

Structural determinants of limited proteolysis.

Limited or regulatory proteolysis plays a critical role in many important biological pathways like blood coagulation, cell proliferation, and apoptosis. A better understanding of mechanisms that control this process is required for discovering new proteolytic events and for developing inhibitors with potential therapeutic value. Two features that determine the susceptibility of peptide bonds to proteolysis are the sequence in the vicinity of the scissile bond and the structural context in which the bond is displayed. In this study, we assessed statistical significance and predictive power of individual structural descriptors and combination thereof for the identification of cleavage sites. The analysis was performed on a data set of >200 proteolytic events documented in CutDB for a variety of mammalian regulatory proteases and their physiological substrates with known 3D structures. The results confirmed the significance and provided a ranking within three main categories of structural features: exposure > flexibility > local interactions. Among secondary structure elements, the largest frequency of proteolytic cleavage was confirmed for loops and lower but significant frequency for helices. Limited proteolysis has lower albeit appreciable frequency of occurrence in certain types of ß-strands, which is in contrast with some previous reports. Descriptors deduced directly from the amino acid sequence displayed only marginal predictive capabilities. Homology-based structural models showed a predictive performance comparable to protein substrates with experimentally established structures. Overall, this study provided a foundation for accurate automated prediction of segments of protein structure susceptible to proteolytic processing and, potentially, other post-translational modifications.

Effects of DMSO on gene expression in human and rat hepatocytes.

Dimethyl sulfoxide (DMSO) is a very common organic solvent used for dissolving lipophilic substances, for example for in vitro cell-based assays. At the same time, DMSO is known to be cytotoxic at high concentrations. Therefore, it is important to define threshold concentrations of DMSO for cells but relevant data at the molecular level are very limited. We have focused on conducting microarray analyses of human and rat hepatocytes treated with more than 100 chemicals in attempts to identify candidate biomarker genes. In the present study, the effects of DMSO on gene expression and cytotoxicity were assessed in human cryopreserved hepatocytes and rat primary cultured hepatocytes. A cytotoxicity test with lactate dehydrogenase (LDH) activity demonstrated DMSO to be noncytotoxic up to a concentration of 2% (v/v) in both cases and there were only few effects on the gene expression profiles up to 0.5% (v/v). The observed differences from controls were considered to be of little toxicological importance, but still need to be taken into account in interpretation of findings when DMSO is used at high concentration.

PMAP: databases for analyzing proteolytic events and pathways.

The Proteolysis MAP (PMAP, http://www.proteolysis.org) is a user-friendly website intended to aid the scientific community in reasoning about proteolytic networks and pathways. PMAP is comprised of five databases, linked together in one environment. The foundation databases, ProteaseDB and SubstrateDB, are driven by an automated annotation pipeline that generates dynamic 'Molecule Pages', rich in molecular information. PMAP also contains two community annotated databases focused on function; CutDB has information on more than 5000 proteolytic events, and ProfileDB is dedicated to information of the substrate recognition specificity of proteases. Together, the content within these four databases will ultimately feed PathwayDB, which will be comprised of known pathways whose function can be dynamically modeled in a rule-based manner, and hypothetical pathways suggested by semi-automated culling of the literature. A Protease Toolkit is also available for the analysis of proteases and proteolysis. Here, we describe how the databases of PMAP can be used to foster understanding of proteolytic pathways, and equally as significant, to reason about proteolysis.