Alzheimer's disease (AD) is a serious neurodegenerative disease consisting of four stages where the illness gets progressively worse. It is of great significance to detect the gene regulatory mechanism as AD progresses and, thus, to help us better understand the causes of AD and find ways to treat or control AD. There are numerous researches to conduct this kind of study. However, the majority of methods are processing region by region of brain, stage by stage of AD, and then compare the results to detect changes. It is unclear how to combine these three dimensions, i.e., gene, region and stage, simultaneously to study gene expression dynamics of AD. This is the motivation of our research. In our study, we propose a statistical model of increments to clarify the relationship between gene expression in adjacent stages, so that we could better estimate the missing data we want and obtain a complete reasonable dynamic regulatory network model. Simulations are conducted to validate the statistical power of our algorithm. Moreover, a real data analysis shows that our method can capture the dynamic gene regulatory relationships among this complex brain data.

翻译：阿尔茨海默症（AD）是一种严重的神经退行性疾病，包含四个病情逐步恶化的阶段。揭示AD进展过程中的基因调控机制，对于理解AD病因并寻找治疗或控制方法具有重要意义。现有大量研究致力于此类探索，然而大多数方法对大脑区域和AD阶段进行逐区逐阶段分析后，再通过结果比较检测变化。如何同时整合基因、脑区和疾病阶段这三个维度来研究AD的基因表达动态尚不明确。这正是本研究的出发点。我们提出一种增量统计模型，以阐明相邻阶段基因表达之间的关系，从而更准确地估计缺失数据，获得完整合理的动态调控网络模型。通过模拟实验验证了算法的统计效能，真实数据分析表明，该方法能够捕捉复杂脑数据中的动态基因调控关系。