基于稀疏表示变量选择方法的影像遗传学数据分析

目的:采用影像遗传学研究方法探索精神分裂症的影像遗传学特征。方法:在传统稀疏回归模型的基础上,改进了其在不同范数条件下进行变量选择的能力,形成一种基于稀疏表示变量选择算法,并将该算法应用于208个受试者的41236个功能磁共振成像数据和722177个单核苷酸多态性数据的综合分析。通过对两类数据施加不同的权重因子,并使用不同的Lp(p=0、0.5、1)范数分别对模型进行求解,筛选出两类数据在不同条件下的显著特征。结果:基因DAOA和HTR2A在3种范数下均被筛选出。此外,在影像学数据方面,发现中央前回、枕上回、顶下缘角回、角回、内侧和旁扣带脑回、后扣带回脑区与精神分裂症相关,此发现与先前精神分裂症的临床医学研究结果一致。结论:基于稀疏表示变量选择方法应用于影像遗传学数据分析是一个有效可行的途径,为今后精神分裂症的影像遗传学研究提供了一种新的研究思路。

Sparse representation-based variable selection algorithm for analysis of imaging genetics data

Objective To explore the imaging genetic characteristics of schizophrenia using imaging genetics method. Methods A sparse representation-based variable selection algorithm with improved ability of variable selection under different norm conditions based on traditional sparse regression model is proposed. The proposed algorithm was applied for the comprehensive analysis of 41 236 functional magnetic resonance imaging data and 722 177 single nucleotide polymorphisms data of 208 subjects. By applying different weight factors to the two types of data and using different Lp (p=0, 0.5, 1) norms for solving the models, the significant features of the two types of data were extracted. Results DAOA and HTR2A genes were extracted under 3 different Lp norms. In addition, the results of imaging data suggested that precentral, occipital_sup, parietal_inf, angular, cingulum_mid, cingulum_post were associated with schizophrenia, which was consistent with previous clinical studies on schizophrenia. Conclusion Sparse representation-based variable selection algorithm is an effective and feasible approach for the analysis of image genetics data, providing a new direction for the image genetics study on schizophrenia.