Classification Of Bipolar Disorder Based On Machine Learning

Bipolar disorder refers to a type of disease that has both manic episodes and depressive episodes.It is a type of mood disorder characterized by fluctuations in mood and energy.At present,the clinical diagnosis and treatment of the disease is often unsatisfactory,with a high rate of misdiagnosis and recurrence.Its pathogenesis is complex,and recent studies have shown that bipolar disorder is highly heritable,but the specific pathogenesis is still unclear.We downloaded 1868 cases of bipolar disorder and 3000 healthy control samples from WTCCC,each of them contains 469612 SNPs.Firstly we conducted a genome-wide association analysis between a group of bipolar disorder cases and their healthy controls,and screened 164 SNP loci that may be related to the risk of bipolar disorder.Secondly,using the selected SNP as a molecular genetic marker,combined with convolutional neural network to construct a two-phase barrier recognition model.The experimental results show that the overall recognition rate of bipolar disorder and healthy control group can reach to 79%,compared with similar studies,our model obtained excellent performance.Finally,we used the selected SNP loci as characteristic sites to cluster the 1868 bipolar disorder patients,they were clustered 1868 patients into 48 clusters.We furtherly analyzed each cluster,extracted the same combination of SNPs.The relevant SNPs were mapped to genes by the NCBI database,and the GeneCards database was used to query the functions of the corresponding genes.In the process,14 genes directly or indirectly related to bipolar disorder were found.This is the first attempt to use genome-wide association analysis as a feature selection method in machine learning data preprocessing,which is a new exploration of GWAS research application;the results of classification and clustering show that machine learning technology has a certain application space in the study of bipolar disorder.Combining existing diagnosis and treatment with molecular genetic data may contribute to the classification and diagnosis of bipolar disorder,and provide new ideas for the development of targeted drugs for this illness.