Multi-Modality Imaging Genetics Study Of Major Depressive Disorder: Exploring The Relationship Between Genetics And Brain Imaging In MDD

Major Depressive Disorder(MDD)is a common and serious disorder.It is estimated that5% of adults worldwide suffer from this disorder.In recent years,imaging genetics has received increasing attention and has become an important research method for discovering associations between genetic variants,such as single nucleotide polymorphisms(SNPs),and brain imaging data.However,most of the existing MDD imaging genetics studies conducted by clinicians use simple statistical analysis methods that consider only unimodal brain imaging,which limits deeper findings in the understanding of MDD mechanisms.Therefore,it is essential to fully explore the relationship between genetic variants and multimodal brain imaging.In this paper,we use multimodal imaging data from structural magnetic resonance imaging(s MRI)and resting-state functional magnetic resonance imaging(rs-f MRI)to To investigate how to fully characterize and utilize the structural information inherent in biomarker data to build correlation models and analyze the correlation between susceptibility genes and brain structure and function to reveal the underlying mechanisms that generate brain cognitive behaviors or related disorders,and then to conduct research on(1)association analysis of genetic loci and multimodal brain networks based on Laplace regularization combined with multilevel diagnostic information;(2)association analysis of genetic loci and multimodal brain networks based on fusion self Expression networks combined with diagnostic information for locus and multimodal brain network association analysis.The main work and innovation points are as follows:(1)In this study,we developed a new imaging genetic association framework based on Laplace regularization to mine multimodal phenotypic networks between genetic risk variants and multistage diagnostic states.Specifically,the multimodal phenotype network consists of voxel node features for s MRI and brain region connectivity features for rs-f MRI.A correlation model based on a multitask learning strategy was used to comprehensively explore the relationship between MDD risk SNPs and the multimodal phenotype network.In addition,multi-stage diagnostic states were introduced to further explore the relationship between multimodality in different subjects.The dataset in this paper was obtained from Zhongda Hospital affiliated with Southeast University and the Second Hospital affiliated with Xinxiang Medical University,and contained multimodal imaging data and gene sequencing data.The experimental results demonstrated the validity of our proposed method,and the method was able to identify some brain regions of interest(ROIs)closely associated with MDD.In addition,the method identified four new potential risk SNPs associated with MDD.(2)To further explore the association of MDD risk gene SNPs with multimodal imaging data,this study embedded the diagnostic information into the multimodal data of brain regions by fusing the self-expression network before model association.By reconstructing the original data using the self-expression properties,the similar structure of the data can be better described.Specifically,the self-expression network is first constructed by sparse representation using intra-class similarity information.Then,the self-expression networks of different modal brain phenotypes are fused.Finally,association analysis is performed based on the fused self-expression networks.Using the same dataset described above,the experimental results validate that our proposed method not only better estimates the potential associations between genetic markers and quantitative traits.In addition,the method identified 15 new potential risk SNPs associated with MDD.In this thesis,we used machine learning algorithms to fully exploit the association between genetic variants and multimodal brain images to identify ROIs closely associated with MDD,in addition to discovering several potential risk SNPs associated with MDD,and these discovered biomarkers and SNPs can help to further explain and explore the pathogenesis of MDD.