SNP Selection Technology And Design And Implementation Of Schizophrenia Diagnosis System

Schizophrenia is a genetic disease with a high recurrence rate and a long morbidity.It is often seen in young and middle-aged patients,and it brings a lot of burden to patients and families.The Single Nucleotide Polymorphism(SNP)Genome-Wide Association Study(GWAS)has achieved remarkable results in the diagnosis of schizophrenia,but the SNP locus There is more redundancy between them.Therefore,in order to apply SNP data to the diagnosis of complex diseases,it is necessary to select a representative subset of information SNPs.With the development of data analysis and processing technology,researchers can use machine learning to mine disease pathogenesis and design diagnostic models from a large amount of data.This thesis takes schizophrenia as the research object,and mainly discusses the selection method of SNP locus and the construction of diagnostic model.Firstly,an information SNP subset selection method based on improved ant colony algorithm is proposed.A representative feature SNP subset is selected from all SNPs to reduce redundancy and noise information.Then a random forest based schizophrenia diagnosis model is designed.Finally Design and implement a schizophrenia diagnostic system.This article mainly contains the following three contents:(1)For the existing information SNP selection method,SNP data is not considered to reflect the characteristics of case SNP data.This study proposes a new feature selection method based on ant colony algorithm and applies it to the selection of SNP.In this thesis,the unique linkage disequilibrium of SNP data is introduced into the heuristic function of ant colony algorithm,and the heuristic function is reconstructed.At the same time,the pheromone update mechanism is redesigned,which can prevent the algorithm from falling into local optimum by adjusting the pheromone volatilization speed.The experimental results show that compared with other selection algorithms,the newly proposed ant colony method constructs a subset of information SNPs with better reconstruction accuracy for non-information SNPs,and the average accuracy of classification experiments is higher than other methods.It was 2.31% and 3.46%.Therefore,the proposed SNP selection method based on improved ant colony has advantages in the selection of information SNP subsets.(2)For the traditional random forest algorithm,when selecting data and features,different degrees of influence of different SNP data and features on schizophrenia are not considered.A random forest algorithm based on weight fusion is proposed.The first is the data weight.The basic idea is that the SNP data of schizophrenia patients has the same number of sites as other patient data,and the fewer the SNP data with healthy people,the greater the weight of the disease data,health data.The same reason.Then a new feature evaluation method is proposed,and the newly proposed calculation method is combined with the chi-square test and the ReliefF algorithm to determine the feature weights respectively.The weight coefficients of the three evaluation criteria are determined according to the order relationship analysis method.Finally,according to the fusion model.The obtained feature weight results are differentiated random selection features in the feature space.The experimental results show that the proposed algorithm has great advantages in calculating generalization error and classification accuracy.(3)Based on the above research,this thesis also completed the design and implementation of intelligent diagnosis prototype system based on SNP data for schizophrenia,including information SNP selection and random forest model construction.Tests show that the system improves the accuracy of SNP selection and the correctness of classification diagnosis to some extent.