Methods And Tools For Integrated Analysis Of Multi-omics Data Towards Assisted Diagnosis And Treatment To Serious Diseases

Methods and tools for integrated analysis of multi-omics datatowards assisted diagnosis and treatment to serious diseasesRecently,infectious diseases,chemical agents and error use of synthetic biotechnology results in rapid growth of public health emergencies.The existing technology cannot effectively deal with the unknown public health emergencies.Besides,the diagnosis and treatment for serious diseases is still limited to single omics data.Integrating multi-omics data for diagnosis and treatment for serious diseases and drug research is still on the way.In recent years,the international large-scale scientific program has accumulated amounts of multi-omics data of genetic perturbation,drug stimulation and disease,resulting in rapid development of system biology.It not only provides the opportunity to explore a new generation of public health emergencies system based on big data,also offers us rich data source and methodology to assist the diagnosis and treatment of serious diseases.This research focuses on the integrated analysis of multi-omics data for assisted diagnosis and treatment of serious diseases.In the part of anti-viral drug repositioning based on knowledge mining and association network,this research systematically studied association between virus and host protein,and predicted anti-viral drugs.Firstly,we divided viruses into four classes,and discovered the similarities and differences of four virus classes in the aspects of transcriptional regulation,PPI network and essential genes.Secondly,using known drug-target interactions,we systematically evaluated the potential of virus host factors to be repositioned for drug targets.Finally,we repositioned approved drugs for anti-HSV,anti-IAV-H1N1,and anti-HBV,and validated the availability of predicted drugs.The experimental validation showed that bosutinib and dexamethasone had certain depressant effect against HSV-1;dextromethorphan and atomoxetine performed inhibited effect in a low dose of IAV-H1N1;melatonin,resveratrol,and etodolac could indeed inhibit the HBV.This research established the system of rapid discovery of biochemical agent's damage mechanisms and rapid drug repositioning based on transcriptional data.We systematically collected extensive transcriptional data for known biochemical agent's damage,genetic perturbation and drug stimulation from the GEO and LINCS database,and preprocessed the data to construct the local data sets.After that,we used the GSEA method to compare the cellular response data of biochemical agent with those of genetic and drug perturbation.Using community mining and clustering analysis with the known biochemical database,we uncovered the biochemical agent's damage mechanisms and predicted drugs for treatment.Then,we used the Softmax and DNN model to discover new indications and new target proteins for all approved drugs in the LINCS database.This research also established several analysis framework based on multi-omics data fusion.Firstly,by integrating three common types of integrated clustering methods,we designed a web server ICM for integrated clustering of multi-dimensional biomedical data.Moreover,we took the classification of LAML as example to illustrate the capacity of ICM in the aspect of assisted diagnosis and treatment for serious diseases.Secondly,we proposed two algorithms for multi-dimensional data fusion based on random walk with restart,named RWRF and RWRNF,and took six types of tumor as example to illustrate the strengths of new methods.Thirdly,we developed an integrative approach for drug repositioning using multiple characterization fusion,named PIMD,and reported the construction of an integrated drug similarity network.Through five kinds of statistical analysis,we repositioned approved drugs from two aspects.Finally,by integrating multi-omics data of drug,target and disease,we proposed a compound-target-disease association prediction framework based on multi-dimensional data fusion,named PAMDF.We used ten-fold cross-validation,biological database cross-validation and community analysis to validate the reliability of PAMDF.The result illustrated that PAMDF is better than prediction methods based on single omics data or other algorithms for binary association prediction.Lastly,the research was summarized and further possible research was prospected.The main innovations of this research include the following two points.Firstly,compared with traditional strategy for public health emergencies,the association knowledge and transcriptional data-supported system is economically attractive,low risk and time saving.With the accumulation of public data from labs around the world,the system is highly extensible.Secondly,we proposed and established a series of platform,algorithm and framework for multi-omics data fusion.By applying the strategy of multi-omics data fusion analysis,researchers can investigate the development of serious diseases and the mechanisms of drugs from multiple perspective and deal with unknown public health emergencies rapider and better.