联合脑成像网络与基因分析预测阿尔茨海默症

Dementia, such as Alzheimer’s disease (AD), are becoming increasing health burdens for human societies. This is a challenging and important research area that can greatly improve the quality of human life. Currently, there are no effective methods to diagnose early AD and treat for the disease. In regards to genetic research on AD, several genes associated with potential development of AD have been found. Our previous studies have shown complex multi single nucleotide polymorphisms (SNPs) and multi quantitative trait (QT) associations for AD. The latest research demonstrates that network topology and specific biologic markers for AD can both help with early detection and also finding new drug targets to improve medication efficacy. However, the integration of neural networks with genetic analysis in AD research has not been established. This project is to develop a multi-level brain connectome analysis through the construction of a sophisticated multi-model brain network that is capable of deciphering and analyzing the network changes in neurological diseases. The goal is to find different imaging biomarkers for neurological diseases at the network level. The associations between genetics, phenotypic changes, and differences in brain network connections will be investigated to identify specific characteristics of AD. An evaluation system of comprehensive influence factor that combines both the genetic sensitivity and neuroimaging specificity will be utilized to predict AD at higher correlation coefficient and lower error. Our project integrates both genetic testing and neuroimaging techniques to study and identify early AD biomarkers, which has significant translational value in the clinical setting.

与老龄化相关的阿尔茨海默症等脑病日益成为人类的巨大健康负担，这是一个可以大幅提高人类生存质量的极其重要的研究领域。目前还没有任何有效的早期诊断方法和治疗手段。有关的基因学研究，已成功的检测出多个相关的风险基因，我们前期关于脑成像基因学的研究，也已揭示了有复杂的多重单核苷酸多态性与多重数量性状间的关联。近年来神经网络技术发展迅速，有研究发现网络拓扑学和生物学标记都能帮助早期检测和改进药物研制，但融合神经网络与基因学分析的研究尚未深入开展。本项目基于更精细、客观的大脑分区构建脑神经网络，旨在建立多层次的脑连接的计算理论和脑网络分析方法,利用与阿尔茨海默症有显著关联的脑区网络拓扑学影像表征的特异性和风险基因集群的敏感性，集成基因测试和神经网络技术，建立一个对阿尔茨海默症的综合影响因素评估体系，进而提高预测阿尔茨海默症的准确率，并推动其在临床上的应用。

中国将“脑科学与类脑研究”上升为国家战略意图，以探索大脑认知的神经原理为主体，研发脑重大疾病诊治新手段和脑机智能新技术为导向的脑科学研究。本项目围绕研究人脑结构性及功能性连接，探索各脑区间的相互作用关系。我们提出的J-HCPMMP方法，首次实现了人脑连接组计划多模态脑分区方法（HCP MMP）在非HCP协议磁共振数据上的应用，实现了精准的脑分区。提出的构建大脑影像连接组可全面而细致地刻画大脑内部的组织模式，即是由彼此纵横交叉相互连接的复杂网络系统。通过研究分析结构和功能性脑网络拓扑属性的差异和统计特性，以揭示阿尔茨海默症在网络层面的变化规律，完善脑网络分析方法，在神经网络水平上发现脑疾病影像学特征。我们研究可导致阿尔茨海默症的风险基因在脑神经网络的表达，探讨基因变异和脑区形态及功能改变之间的关联，及与阿尔茨海默症有显著关联的脑区网络拓扑学影像特异性表征和风险基因集群的敏感度。通过高风险基因异化与表型的关系，明确不同阿尔茨海默症特异性表征，研究不同疾病阶段基因组人脑退化在多模态影像上的差异。综合基因测试和神经网络技术，我们提出了集成费希尔(Fisher)得分和多模态多任务特征选择新方法，建立对阿尔茨海默症的综合影响评估体系，以实现阿尔茨海默症早期诊断和早期预警。本项目丰硕的研究成果对于国内外同行的工作将会大有裨益，也为我们将来的工作打下坚实的基础。