Big Data Based Genome Wide Gene-ADR Network Association Analysis And ADR Prediction

The adverse drug reaction(ADR)is a severe clinical pharmacology and toxicology problem that has caused immense medical and social burdens.It is also one of the two major causes of current new drug discovery failure.Unfortunately,due to the limitation of current experimental technologies,there misses a reproducible method for aiding systematical profiling and mechanistic understanding of most common ADRs.Adverse Drug Reaction Classification System(ADReCS)is a comprehensive ADR ontology database that provides not only ADR standardization but also hierarchical classification of ADR terms.The ADR terms were pre-assigned with unique digital IDs and at the same time were well organized into a four-level ADR hierarchy tree for building an ADR-ADR relation.The ADReCS can serve as the fundamental data source for systems computational toxicity search.In this study,we updated the ADReCS web service in both aspects of data size and novel functions.Comparing to its previous version,the ADReCS adds 1,044 more drugs,303 more standard ADR terms,and 1,433 ADR-protein relations.For the first time,we incorporated information of ADR severity grades into the database,which provides a weight for differentiating ADRs.The latest version of database reposits 6,847 standard ADR terms and 61,113 synonyms,covering single active ingredient 2,399 drug and 202,479 drug-ADR relations.Besides,we deployed novel tools into the ADReCS web service,for instance,search of drugs based on ADR similarity or structure similarity.To aid systems toxicity study via drug-gene regulation,we mined the differentially expressed genes(DEGs)from big data of the drug-treated human cell transcriptomes derived from the Connectivity Map(CMap).We examined the DEGs achived under combination conditions of cell types,drugs,drug treatment concentration and drug exposure time.We found that drug-induced gene expression perturbation was comparatively stronger in the organ systems of their indication sites than that of in other organs.The antineoplastic and immunomodulating agents can differentiate expression of many genes;however,the expression change of every single gene is weak.Furthermore,we demonstrated association analyses between drug-regulated genes and drug-induced ADRs.The results found that similar ADR-associated genes intended to cause similar ADRs in the same organ system.Our findings are of value in aiding assessment and molecular mechanism study of ADRs.In this study,we also developed an improved Naive Bayesian Network Model for high throughput drug safety profiling by solving complex network of drug-gene-ADR interactions.The model can de novo evaluate a drug in a broad spectrum of 1,156 distinct standardized ADRs with estimated frequency and severity grade.For the first time,we summarized the overall drug safety using a toxicity score.Furthermore,we deployed the model as a web service,Adverse Drug Reaction Alert-gene(ADRAlert-gene,http://bioinf.xmu.edu.cn/ADRAlert/gene).In addition,we determined the association strength for each of 3,807,631 gene-ADR pairs.Uing these gene-ADR assoication data,we summarized a network-based strategy in systematic drug toxicology research which can be freely applied in molecular understanding various ADRs.In an example,we identified six potential targets for monitoring or countermining allergic conditions.In summary,this study offers new insights into prioritizing clinical pharmacogenetic tests for safe drug therapy,and it enhances the success rate of new drug discovery by reducing the drug safety problems in clinical trials.It also lies the foundation for the future individual drug therapy.