Ryusuke Sawada

DINIES: drug–target interaction network inference engine based on supervised analysis

DINIES (drug–target interaction network inference engine based on supervised analysis) is a web server for predicting unknown drug–target interaction networks from various types of biological data (e.g. chemical structures, drug side effects, amino acid sequences and protein domains) in the framework of supervised network inference. The originality of DINIES lies in prediction with state-of-the-art machine learning methods, in the integration of heterogeneous biological data and in compatibility with the KEGG database. The DINIES server accepts any ‘profiles’ or precalculated similarity matrices (or ‘kernels’) of drugs and target proteins in tab-delimited file format. When a training data set is submitted to learn a predictive model, users can select either known interaction information in the KEGG DRUG database or their own interaction data. The user can also select an algorithm for supervised network inference, select various parameters in the method and specify weights for heterogeneous data integration. The server can provide integrative analyses with useful components in KEGG, such as biological pathways, functional hierarchy and human diseases. DINIES (http://www.genome.jp/tools/dinies/) is publicly available as one of the genome analysis tools in GenomeNet.

Systematic drug repositioning for a wide range of diseases with integrative analyses of phenotypic and molecular data.

Drug repositioning, or the application of known drugs to new indications, is a challenging issue in pharmaceutical science. In this study, we developed a new computational method to predict unknown drug indications for systematic drug repositioning in a framework of supervised network inference. We defined a descriptor for each drug-disease pair based on the phenotypic features of drugs (e.g., medicinal effects and side effects) and various molecular features of diseases (e.g., disease-causing genes, diagnostic markers, disease-related pathways, and environmental factors) and constructed a statistical model to predict new drug-disease associations for a wide range of diseases in the International Classification of Diseases. Our results show that the proposed method outperforms previous methods in terms of accuracy and applicability, and its performance does not depend on drug chemical structure similarity. Finally, we performed a comprehensive prediction of a drug-disease association network consisting of 2349 drugs and 858 diseases and described biologically meaningful examples of newly predicted drug indications for several types of cancers and nonhereditary diseases.

Benchmarking a Wide Range of Chemical Descriptors for Drug‐Target Interaction Prediction Using a Chemogenomic Approach

The identification of drug‐target interactions, or interactions between drug candidate compounds and target candidate proteins, is a crucial process in genomic drug discovery. In silico chemogenomic methods are recently recognized as a promising approach for genome‐wide scale prediction of drug‐target interactions, but the prediction performance depends heavily on the descriptors and similarity measures of drugs and proteins. In this paper, we investigated the performance of various descriptors and similarity measures of drugs and proteins for the drug‐target interaction prediction using a chemogenomic approach. We compared the prediction accuracy of 18 chemical descriptors of drugs (e.g., ECFP, FCFP,E‐state, CDK, KlekotaRoth, MACCS, PubChem, Dragon, KCF‐S, and graph kernels) and 4 descriptors of proteins (e.g., amino acid composition, domain profile, local sequence similarity, and string kernel) on about one hundred thousand drug‐target interactions. We examined the combinatorial effects of drug descriptors and protein descriptors using the same benchmark data under several experimental conditions. Large‐scale experiments showed that our proposed KCF‐S descriptor worked the best in terms of prediction accuracy. The comparative results are expected to be useful for selecting chemical descriptors in various pharmaceutical applications.

Target-Based Drug Repositioning Using Large-Scale Chemical-Protein Interactome Data

Drug repositioning, or the identification of new indications for known drugs, is a useful strategy for drug discovery. In this study, we developed novel computational methods to predict potential drug targets and new drug indications for systematic drug repositioning using large-scale chemical-protein interactome data. We explored the target space of drugs (including primary targets and off-targets) based on chemical structure similarity and phenotypic effect similarity by making optimal use of millions of compound-protein interactions. On the basis of the target profiles of drugs, we constructed statistical models to predict new drug indications for a wide range of diseases with various molecular features. The proposed method outperformed previous methods in terms of interpretability, applicability, and accuracy. Finally, we conducted a comprehensive prediction of the drug-target-disease association network for 8270 drugs and 1401 diseases and showed biologically meaningful examples of newly predicted drug targets and drug indications. The predictive model is useful to understand the mechanisms of the predicted drug indications.

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

BackgroundPhenotype-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.MethodsIn 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.”ResultsUnlike 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.ConclusionsThe transcriptomic approach reported here is expected to be a useful tool for structure-independent prediction of target proteins for drug candidate compounds.

Elucidating the modes of action for bioactive compounds in a cell-specific manner by large-scale chemically-induced transcriptomics

The identification of the modes of action of bioactive compounds is a major challenge in chemical systems biology of diseases. Genome-wide expression profiling of transcriptional responses to compound treatment for human cell lines is a promising unbiased approach for the mode-of-action analysis. Here we developed a novel approach to elucidate the modes of action of bioactive compounds in a cell-specific manner using large-scale chemically-induced transcriptome data acquired from the Library of Integrated Network-based Cellular Signatures (LINCS), and analyzed 16,268 compounds and 68 human cell lines. First, we performed pathway enrichment analyses of regulated genes to reveal active pathways among 163 biological pathways. Next, we explored potential target proteins (including primary targets and off-targets) with cell-specific transcriptional similarity using chemical–protein interactome. Finally, we predicted new therapeutic indications for 461 diseases based on the target proteins. We showed the usefulness of the proposed approach in terms of prediction coverage, interpretation, and large-scale applicability, and validated the new prediction results experimentally by an in vitro cellular assay. The approach has a high potential for advancing drug discovery and repositioning.

Large-Scale Prediction of Beneficial Drug Combinations Using Drug Efficacy and Target Profiles

The identification of beneficial drug combinations is a challenging issue in pharmaceutical and clinical research toward combinatorial drug therapy. In the present study, we developed a novel computational method for large-scale prediction of beneficial drug combinations using drug efficacy and target profiles. We designed an informative descriptor for each drug-drug pair based on multiple drug profiles representing drug-targeted proteins and Anatomical Therapeutic Chemical Classification System codes. Then, we constructed a predictive model by learning a sparsity-induced classifier based on known drug combinations from the Orange Book and KEGG DRUG databases. Our results show that the proposed method outperforms the previous methods in terms of the accuracy of high-confidence predictions, and the extracted features are biologically meaningful. Finally, we performed a comprehensive prediction of novel drug combinations for 2,639 approved drugs, which predicted 142,988 new potentially beneficial drug-drug pairs. We showed several examples of successfully predicted drug combinations for a variety of diseases.

What parameters characterize “life”?

“Life” is a particular state of matter, and matter is composed of various molecules. The state corresponding to “life” is ultimately determined by the genome sequence, and this sequence determines the conditions necessary for survival of the organism. In order to elucidate one parameter characterizing the state of “life”, we analyzed the amino acid sequences encoded in the total genomes of 557 prokaryotes and 40 eukaryotes using a membrane protein prediction online tool called SOSUI. SOSUI uses only the physical parameters of the encoded amino acid sequences to make its predictions. The ratio of membrane proteins in a genome predicted by the SOSUI online tool was around 23% for all genomes, indicating that this parameter is controlled by some mechanism in cells. In order to identify the property of genome DNA sequences that is the possible cause of the constant ratio of membrane proteins, we analyzed the nucleotide compositions at codon positions and observed the existence of systematic biases distinct from those expected based on random distribution. We hypothesize that the constant ratio of membrane proteins is the result of random mutations restricted by the systematic biases inherent to nucleotide codon composition. A new approach to the biological sciences based on the holistic analysis of whole genomes is discussed in order to elucidate the principles underlying “life” at the biological system level.

Predicting inhibitory and activatory drug targets by chemically and genetically perturbed transcriptome signatures

Genome-wide identification of all target proteins of drug candidate compounds is a challenging issue in drug discovery. Moreover, emerging phenotypic effects, including therapeutic and adverse effects, are heavily dependent on the inhibition or activation of target proteins. Here we propose a novel computational method for predicting inhibitory and activatory targets of drug candidate compounds. Specifically, we integrated chemically-induced and genetically-perturbed gene expression profiles in human cell lines, which avoided dependence on chemical structures of compounds or proteins. Predictive models for individual target proteins were simultaneously constructed by the joint learning algorithm based on transcriptomic changes in global patterns of gene expression profiles following chemical treatments, and following knock-down and over-expression of proteins. This method discriminates between inhibitory and activatory targets and enables accurate identification of therapeutic effects. Herein, we comprehensively predicted drug–target–disease association networks for 1,124 drugs, 829 target proteins, and 365 human diseases, and validated some of these predictions in vitro. The proposed method is expected to facilitate identification of new drug indications and potential adverse effects.

KampoDB, database of predicted targets and functional annotations of natural medicines

Natural medicines (i.e., herbal medicines, traditional formulas) are useful for treatment of multifactorial and chronic diseases. Here, we present KampoDB (http://wakanmoview.inm.u-toyama.ac.jp/kampo/), a novel platform for the analysis of natural medicines, which provides various useful scientific resources on Japanese traditional formulas Kampo medicines, constituent herbal drugs, constituent compounds, and target proteins of these constituent compounds. Potential target proteins of these constituent compounds were predicted by docking simulations and machine learning methods based on large-scale omics data (e.g., genome, proteome, metabolome, interactome). The current version of KampoDB contains 42 Kampo medicines, 54 crude drugs, 1230 constituent compounds, 460 known target proteins, and 1369 potential target proteins, and has functional annotations for biological pathways and molecular functions. KampoDB is useful for mode-of-action analysis of natural medicines and prediction of new indications for a wide range of diseases.