Exploitation Of Voxel-wise GWAS Based On Spatial Information For Biomarker Detection Of AD

Important biomarker detection of alzheimer’s disease(AD)is crucial for prediction,diagnosis,and monitoring of AD.Many methods have been proposed to detect potential biomarkers of AD in imaging genomics studies,the voxel-wise genome-wide association analysis(vGWAS)method is one of common method.However,it is challenging to perform vGWAS because of its computational complexity.Furthermore,owing to treating each voxel as an independent unit,vGWAS ignorance of the spatial information of images,which could lead to some false results or miss some important biomarker.Thus,when perform the traditional vGWAS,it would face problematic.In order to address these problems,in this paper,we proposed a novel method that may help detecting important biomarkers of AD by using the accelerate procedure of vGWAS--fast voxel-wise genome-wide association analysis(FVGWAS)--based on the exploitation of spatial information.Which not only address the problem of computational complexity,but also improve the powder of detect the biomarkers of AD.To incorporate spatial information in vGWAS,we applied three commonly used methods for image processing,including gaussian weight(GS),Non-Local Means(NLM)weight,and BM3D/BM4D(block-matching with 3D filtering/block-matching with 4D filtering)weight.Subsequently,the images were analyzed with the genotype data using FVGWAS procedure for test the important biomarker of the brain images and genotype data.The proposed method was estimated on both simulated and real data(obtain from ADNI dataset).For simulated data,experiments show that incorporating spatial information can improve the estimated accuracy.Moreover,the Gaussian weight procedure has the best results among the three methods.For real data,which obtained from ADNI,our method identified three genes,ANK3,MEIS2,and TLR4,that demonstrate significant associations with mental retardation and learning disabilities and that are related to age,which suggests that these genes could play a role in influencing AD or other neurodegenerative diseases.We also found some significant clusters,including the hippocampus,the inferior frontal gyrus,the insula,and the precuneus at al,which have been demonstrate associate with memory,language comprehension,and cognitive function,thus,those cluster may be strong connective with AD or other neurodegenerative diseases.Our results demonstrate that the proposed method may be an effective and valuable tool to detect potential biomarkers of AD in imaging phenotype data and genotype data.