Establishment And Application Of Network Entropy Technology For Complex Diseases

With rapid development of multiomics-based high throughput technology,biologists have identified a large number of disease-candidate genes.These genes are related to occurrence,development and treatment of diseases.However,due to possible noise and sample bias in acquisition of these data,there is often a large amount of redundant information in existing disease-candidate genes,which may hinder our understanding of molecular mechanism of diseases.Therefore,quantitative characterizing regulation abilities of diseasecandidate genes(DCG)in different disease environments and prioritizing these genes are crucial for understanding disease progression and developing related high-efficiency drugs.Traditional methods for finding master genes of different diseases often focus on the constant global protein interaction network,ignoring the fact that intermolecular interactions change among different disease contexts.What's more,the same gene may play completely different functions in different environments.Therefore,in this study,based on the chaos theory of complex disease systems,we proposed a concept of network-based disease entropy,which is used to characterize the regulation abilities of disease-candidate genes in different disease environments,and systematically explore the co-occurrence of diseases,drug repositioning,and mechanism of traditional Chinese medicines.Firstly,based on a directed protein interaction network,we constructed specific regulatory networks for 274 diseases to simulate their specific molecular environment and build the disease network entropy theory.We then proposed the Network-Oriented Gene Entropy Approach(NOGEA)to rank the regulatory ability of each DCG and infer master genes.Using the genetic data related to disease occurrence and development in existing databases,we found that the master genes inferred by NOGEA were involved in the occurrence and development of the disease,indicating the reliability of NOGEA.Compared with other DCG prioritizing algorithms,we further confirmed that NOGEA is a reliable approach.Elucidating molecular mechanisms of disease co-occurrence play an important role in understanding disease development.To further demonstrate the reliability of disease network entropy theory,we classified DCGs into three types,i.e.“Master”,“Interim”,and “Redundant”.By comparing with the human disease comorbidity network(HDCN),we found that the disease-disease network constructed with master genes exhibit the best fit with HDCN,even better than the disease-disease network constructed using all DCGs.This result indicates that prioritizing disease genes using gene entropy values can reduce noise contained in DCGs to some extent.We then found that interactions between master genes for different diseases can explain the mechanisms of disease co-occurrence.Finally,we used the master genes to explore the potential mechanisms of alcoholism leading to Parkinson's disease.To investigate the application of disease network entropy theory in drug repurposing,we proposed the concept of drug perturbation entropy(DDE)to characterize the potential therapeutic effect of each drug on a particular disease.We found that through application of DDE,it is possible to accurately identify known relationships from a large number of random drug-disease relationships.Moreover,its accuracy is significantly higher than several commonly used methods for predicting drug-disease relationships.Using this approach,we explored the potential mechanism of action of several drugs for the pancreatic cancer treatment,and found that highly effective drugs are more likely to target local modules of master genes on the interaction network.We further screened 11 old drugs that may be used to treat pancreatic cancer from FDA-approved drugs,and verified them by in vitro anti-pancreatic cancer cell experiments.Through RNA expression analysis,we also proved our hypothesis about the local module of drug targeting master genes.Finally,based on the disease network entropy theory and the systemic pharmacology analysis framework,we identified the herb Agrimonia pilosa,which is potentially used for anti-pancreatic cancer,and analyzed its potential molecular mechanism.We found that the main active ingredients of Agrimonia pilosa including quercetin,luteoloside,and phenanthrene,may inhibit proliferation of pancreatic cancer cells by regulating the metabolic processes and growth factors of pancreatic cancer.By controlling chemokines and neurotrophic factors signaling pathways,Agrimonia pilosa may block pancreatic cancer invading to the nervous system.In addition,the active ingredients in Agrimony pilosa also have a certain effect on immunoregulation.These results suggest that Agrimony pilosa may achieve the goal of treating pancreatic cancer by reprograming the metabolic and immune microenvironment of pancreatic cancer at different stages of its development.In summary,this method is based on the disorder of disease molecular system,providing a new strategy for mining disease-specific master genes.Its successful application in exploring disease co-occurrence,drug repurposing,and mechanism of action of TCMs has provided novel perspective for complex diseases related research.