Li Xingyi, Li Min, Zheng Ruiqing, Chen Xiang, Xiang Ju, Wu Fang-Xiang, Wang Jianxin
School of Computer Science and Engineering, Central South University, Changsha, China.
Neuroscience Research Center & Department of Basic Medical Sciences, Changsha Medical University, Changsha, China.
Front Genet. 2020 Feb 5;10:1401. doi: 10.3389/fgene.2019.01401. eCollection 2019.
Since similar complex diseases are much alike in clinical symptoms, patients are easily misdiagnosed and mistreated. It is crucial to accurately predict the disease status and identify markers with high sensitivity and specificity for classifying similar complex diseases. Many approaches incorporating network information have been put forward to predict outcomes, but they are not robust because of their low reproducibility. Several pathway-based methods are robust and functionally interpretable. However, few methods characterize the disease-specific states of single samples from the perspective of pathways. In this study, we propose a novel framework, Pathway Activation for Single Sample (PASS), which utilizes the pathway information in a single sample way to better recognize the differences between two similar complex diseases. PASS can mainly be divided into two parts: for each pathway, the extent of perturbation of edges and the statistic difference of genes caused by a single disease sample are quantified; then, a novel method, named as an AUCpath, is applied to evaluate the pathway activation for single samples from the perspective of genes and their interactions. We have applied PASS to two main types of inflammatory bowel disease (IBD) and widely verified the characteristics of PASS. For a new patient, PASS features can be used as the indicators or potential pathway biomarkers to precisely diagnose complex diseases, discover significant features with interpretability and explore changes in the biological mechanisms of diseases.
由于相似的复杂疾病在临床症状上非常相似，患者很容易被误诊和误治。准确预测疾病状态并识别具有高灵敏度和特异性的标志物以对相似的复杂疾病进行分类至关重要。已经提出了许多结合网络信息的方法来预测结果，但由于其低重现性，这些方法并不稳健。几种基于通路的方法是稳健的且具有功能可解释性。然而，很少有方法从通路的角度来表征单个样本的疾病特异性状态。在本研究中，我们提出了一种新颖的框架，即单样本通路激活（PASS），它以单样本的方式利用通路信息来更好地识别两种相似复杂疾病之间的差异。PASS主要可分为两部分：对于每条通路，量化由单个疾病样本引起的边的扰动程度和基因的统计差异；然后，应用一种名为AUCpath的新方法从基因及其相互作用的角度评估单样本的通路激活。我们已将PASS应用于两种主要类型的炎症性肠病（IBD），并广泛验证了PASS的特征。对于新患者，PASS特征可作为指标或潜在的通路生物标志物，以精确诊断复杂疾病、发现具有可解释性的显著特征并探索疾病生物机制的变化。
