A Study On Drug Knowledge Discovery For Depression Based On Multi-Source Data

ObjectiveWith the accelerating pace of life,people are facing increasing competition and pressure in both their personal lives and work,leading to a rising incidence of depression each year.Therefore,it has become increasingly important to search for effective treatment methods.Additionally,the rapid development of disciplines such as genomics,proteomics,and bioinformatics has made it possible to use computer technology for data mining,enabling the search for more efficient and accurate approaches to drug discovery.Based on these circumstances,the present study focuses on mining knowledge related to depression accumulated over the long term from multiple databases.It constructs a heterogeneous association network based on multi-source data and employs deep learning methods to predict potential associations within the network,aiming to discover drug-related knowledge for depression.Furthermore,the study validates the model’s prediction results through literature verification,aiming to provide highly reliable candidate references for the development of antidepressant drugs and make breakthroughs in terms of cost and time required for drug development.Simultaneously,this research provides new directions and methodological ideas for the study of drug knowledge discovery.ObjectDrug-disease relationship related to depressive disorder.MethodsIn this study,depression-related knowledge accumulated for a long time in multi-source databases in the biomedical field,such as Drug Bank database,OMIM database,and SIDER database,was used as the data source.Relevant drug and disease data were obtained by retrieval according to the study’s requirements.After data preprocessing,drug similarity matrix and disease similarity matrix were constructed based on each attribute feature using similarity measurement methods such as cosine similarity,Jaccard similarity coefficient,and Tanimoto similarity coefficient.The drug similarity matrix and disease similarity matrix were merged using quantile standardization and weighted fusion methods to obtain the final drug similarity matrix_and disease similarity matrix_.Based on the known"drug-disease"correlation matrix retrieved from TTD database,the heterogeneous network of"drug-disease"was constructed,and the correlation feature vector of"drug-disease"was extracted.The deep neural network algorithm in deep learning was used to train,optimize,and evaluate the drug knowledge discovery model,which was used to carry out potential correlation prediction of"drug-disease".The prediction results of the model were verified through text verification to complete the research on drug knowledge discovery for depression.ResultsThe drug knowledge discovery model constructed in this study predicted 395potential antidepressant drugs through analysis and processing of input data,involving known drugs for treating depression,as well as drugs for treating nervous system disorders,antipsychotic drugs,and hypoglycemic drugs.Chlorimipramine ranked first with a probability of 0.897.Through text verification,the model’s prediction performance was good,achieving the expected results of this study.ConclusionThis study integrated and analyzed multi-source data using big data and artificial intelligence technology,which discovered potential antidepressant drugs and expanded the direction of research and development for depression treatment.This method considered multiple dimensions of depression such as its pathogenesis and treatment,providing a new direction for drug development that overcomes traditional research limitations.In addition,the fusion of multi-source data from different fields has improved the accuracy and reliability of drug knowledge discovery.It can provide new methodological ideas for discovering new treatment methods and drugs in other disease areas,thus promoting medical research and clinical practice.