| BackgroundChinese medicines is an important part of the treasure of Traditional Chinese medicine(TCM),and it is also the main material basis and means for TCM to treat diseases.The research on the target spectrum of Chinese medicines is not only the basis of understanding the mechanism of the whole prescription,but also the key to realize the modernization and internationalization of TCM.Chinese medicines can be regarded as the combination of its chemical components,by predicting the target of chemical components,the targets of overall action of Chinese medicines or prescription can be predicted.The target research of chemical components of Chinese medicines mainly includes the biology,physiology and pharmacology experimental methods and computer-aided identification methods.Through the experimental research based on biology,physiology and pharmacology to determine the targets of compounds has a specific advantage,but it is not easy to study on a large scale,and it is difficult to get the target spectrum of Chinese medicines.However,the computer-aided method can be used to conduct this research effectively,and the possible targets of the compounds can be comprehensively screened in the simulated state.At present,the computer-aided target prediction method of Chinese medicines mainly comes from the research ideas of western medicines.Based on the principle of virtual screening,the binding energy of drug ligand and target is calculated to predict the molecular target.However,this process ignores the differences between western medicines and Chinese medicines,and there are three main limitations in the prediction of Chinese medicines target.First,the purpose of target prediction is different.Western drugs are mainly synthetic drugs,which can be designed and synthesized according to the structural characteristics of the target,the purpose of target prediction is mainly to design and develop new drugs.However,the chemical components of Chinese medicines are derived from natural products,and the purpose of their target prediction is to discover and identify new functional components.Therefore,the target prediction of chemical components of Chinese medicines should be closely combined with the detection technology,and the target prediction method based on the detection of Chinese medicines can better reflect the actual role of Chinese medicines.Secondly,there are many unknown chemical components in Chinese medicines,and how to predict the targets of unknown chemical components is also an urgent problem to be solved.Thirdly,Chinese medicines is a unified whole composed of a variety of chemical components,which must be systematically studied in order to predict the action target.This requires us to establish a target prediction method based on the detection of Chinese medicines,so as to meet the purpose of target prediction and target spectrum of Chinese medicines.Therefore,this paper proposes to combine the detection technology of Chinese medicines with the target prediction theory to build a new system to predict the targets of Chinese medicines.The application of electrospray ionization tandem mass spectrometry(MS)in the analysis of Chinese medicines has obtained a lot of research results.At the same time,data mining method has been widely used in the research of targets of Chinese medicines due to its high efficiency,high speed and low consumption.On the basis of MS technology and data mining method,combined with the use of graph database,this paper constructed an integrated process from MS detection to target prediction,so as to achieve rapid and efficient target prediction of Chinese medicines.AimsBased on the current problem of target prediction of Chinese medicines,this paper proposed for the first time to combine MS technology and target prediction theory to construct a target prediction model of Chinese medicines by integrating Chinese medicines,chemical components of Chinese medicines,MS technology,and target prediction.Using the model,the targets of Chinese medicines can be directly predicted by MS fragments,realizing the integration of MS detection technology and target prediction function.What's more,it will be beneficial to provide scientific basis for the research on the functional components and new components discovery of Chinese medicines.MethodsPrediction targets of Chinese medicines is a complicated and arduous work,which is an important part of the core research of Chinese medicines.It is also a prerequisite for revealing the complex action mechanism and new drug research of Chinese medicines on the molecular level.MS is an analyzer for the determination of molecule weight.Due to its high sensitivity,strong specificity and good stability,it is very suitable for the analysis of chemical components of Chinese medicines,and has become an important tool for the study of Chinese medicines.Data mining method has been widely used in drug target prediction because of its high efficiency,high speed and low cost,and has achieved certain results.Based on the basic fact that"compounds with similar targets have similar chemical structures" and the hypothesis that "similar chemical structures will produce similar MS fragments",this paper applied the MS detection technology to the field of target prediction,introduced the concept of MS fragments as the feature of target prediction,and constructed a new target prediction system of the Chinese medicines.The main research contents include:(1)By sorting out target-active small molecule relation data in Chembl database and the chemical components of Chinese medicines from our laboratory,this paper used MS fragmentation simulation technology CFM(Competitive-Fragmentation Modeling)to dissociate the compounds and constructed the graph database of compound-MS fragment-target with Neo4j database platform.(2)The association rules was used to study the correlation between MS fragment combination and targets.Different support thresholds were set according to the differences in the number of target molecules,fragment combination rules of different targets were mined,and the pattern recognition system of targets was constructed by entity grammar system,which was applied to predict the targets of Chinese medicines.(3)Naive Bayesian algorithm theory was applied to construct multi-classification target prediction model based on MS fragment probability,and 10-fold cross-validation was used to evaluate the accuracy of the model.(4)The integrated process from MS detection to target prediction was constructed.The high resolution liquid-mass spectrometer was used to collect and analyze the mass spectrum fragments of Lycii Fructus,and 15 unknown components with higher response intensity were selected.The models established in(2)and(3)were used to predict their targets,and the reliability of the models was checked through database comparison.Results(1)A database of compound MS fragments and Chinese medicines targets was constructed based on ESI tandem MS simulation.By sorting out the target-active small molecule relation data in Chembl database and the chemical components of Chinese medicines-Chinese medicines relation data in our laboratory,CFM technology was used to simulate the MS fragmentation of compounds according to different collision energies,and Neo4j database platform was used to construct a CMFT(Chinese medicines Fragment Target Relationship)database based on Chinese medicines,compounds,characteristic MS fragments of compounds,targets and their mutual relations.This database can quickly query and acquire various nodes and relationships information,and provide a data basis for the next prediction of the action targets of Chinese medicines.(2)A target pattern recognition system based on fragment combination was constructed.Based on the MS fragments of compounds,the association rule algorithm was used to mine fragment combination rules of different targets.Using entity grammar system,a target pattern recognition system based on MS fragment combination was constructed on Neo4j database platform.A total of 917 target modes were obtained under three collision energies,which realized the bidirectional prediction function of Chinese medicines and targets in CMFT database.The prediction of Ardisiae Japonicae Herba showed that the system predicted four confirmed targets,which was suitable for rapid prediction of multi-components and multi-targets of Chinese medicines.(3)A multi-classification target prediction model based on fragment probability was constructed.According to Naive Bayesian algorithm theory,a multi-classification target prediction model based on fragment probability of MS was constructed,which can predict 1026 targets simultaneously.The model was evaluated by 10-fold cross validation,and the overall accuracy was 62.77%.At the same time,the classification ranking concept was introduced,and the 12 targets with the highest probability were predicted as potential targets of the compound with over 80%accuracy.After the application,it can be seen that the model had good prediction results for the compounds within the data set,and also had certain prediction effects for the external compounds.It is suitable for the target prediction study of Chinese medicines and expands the applicability of MS detection technology to target prediction.(4)An integrated process for target prediction based on MS technology was constructed.From the perspective of MS technology,an integrated process from MS technology to target prediction was constructed.Meanwhile,the MS fragments of Lycii Fructus were collected and analyzed by high resolution liquid chromatography-mass spectrometer,15 unknown components with higher response intensity were selected,and their targets were predicted using the model established in(2)and(3).Predicted results were compared with TCM Systems Pharmacology Database and Analysis Platform(TCMSP),the model in(2)predicted five confirmed targets,the model in(3)predicted seven confirmed targets,indicating that MS technology can be used to predict the targets of Chinese medicines,and provide help for the clinical application and new components research of Chinese medicines.ConclusionsIn this study,a CMFT database of active small molecules,targets,Chinese medicines and chemical components of Chinese medicines based on MS fragments was established,and the association between target and fragment was excavated.A target pattern recognition system based on fragment combinations was constructed using entity grammar system,which realized the fast bidirectional prediction of multi-components and multi-targets of Chinese medicines.Meanwhile,using Naive Bayesian theory,a multi-classification target prediction model based on fragment probability was constructed,which can realize the target classification and ranking prediction of more than 1000 targets at the same time.On the other hand,this paper applied MS technology to target prediction for the first time,realized the direct application of MS detection results to target prediction of Chinese medicines,and also accomplished the target prediction of unknown components.The research results of this paper propose a new idea for Chinese medicines research,establish the connection between detection and target prediction of Chinese medicines.It can provide a rapid guidance for the clinical application of Chinese medicines,help to reveal the mechanism of Chinese medicines and promote the research of new components on the molecular level,and also help to accelerate the process of modernization and internationalization of TCM. |
PDF Full Text Request