基于马尔科夫随机场对二型糖尿病早期诊断

A Markov Random Field-Based Approach for Early Diagnosis of Type 2 Diabetes

为了探究二型糖尿病早期动态变化,本文通过分析小鼠标本的组织特异性表达数据,获得相邻时间段之间的差异表达的证据,以表征基因差异表达的动态变化。该数据集中含有丰富的时空信息,但以往的研究中很难充分利用。我们通过在潜在状态上指定马尔可夫随机场(MRF)合并具有时空结构的复杂数据,用蒙特卡洛期望最大化(MCEM)算法计算模型参数,其关键特征是同时考虑基因表达水平的空间相似性和时间依赖性,仿真研究与实例分析结果都表明该方法具有更高的灵敏度,能识别更多的DE基因,能从数据中提取更多生物学上有意义的结果。

In order to explore the early dynamic changes of type 2 diabetes mellitus, this paper analyzed the tissue-specific expression data of mouse specimens to obtain the evidence of differential expression between adjacent time periods, so as to characterize the dynamic changes of differential gene expression. This dataset contains abundant spatiotemporal information, but it is difficult to make full use of it in previous studies. We pass in the potential state specified on Markov random field (MRF) combined with space-time structure of complex data, using the Monte Carlo calculation model for the expectation maximization (MCEM) algorithm parameters, its key characteristic is also considering gene expression level of similarity space and time dependence, simulation research and the example analysis results show that this method has higher sensitivity. More DE genes can be identified and more biologically significant results can be extracted from the data.