Zhang Yi, Chen Min, Cheng Xiaohui, Chen Zheng
School of Information Science and Engineering, Guilin University of Technology 541004 Guilin China.
School of Computer Science and Technology, Hunan Institute of Technology 421002 Hengyang China
RSC Adv. 2019 Sep 20;9(51):29747-29759. doi: 10.1039/c9ra05554a. eCollection 2019 Sep 18.
Lots of research findings have indicated that miRNAs (microRNAs) are involved in many important biological processes; their mutations and disorders are closely related to diseases, therefore, determining the associations between human diseases and miRNAs is key to understand pathogenic mechanisms. Existing biological experimental methods for identifying miRNA-disease associations are usually expensive and time consuming. Therefore, the development of efficient and reliable computational methods for identifying disease-related miRNAs has become an important topic in the field of biological research in recent years. In this study, we developed a novel miRNA-disease association prediction model using a Laplacian score of the graphs and space projection federated method (LSGSP). This integrates experimentally validated miRNA-disease associations, disease semantic similarity scores, miRNA functional scores, and miRNA family information to build a new disease similarity network and miRNA similarity network, and then obtains the global similarities of these networks through calculating the Laplacian score of the graphs, based on which the miRNA-disease weighted network can be constructed through combination with the miRNA-disease Boolean network. Finally, the miRNA-disease score was obtained projecting the miRNA space and disease space onto the miRNA-disease weighted network. Compared with several other state-of-the-art methods, using leave-one-out cross validation (LOOCV) to evaluate the accuracy of LSGSP with respect to a benchmark dataset, prediction dataset and compare dataset, LSGSP showed excellent predictive performance with high AUC values of 0.9221, 0.9745 and 0.9194, respectively. In addition, for prostate neoplasms and lung neoplasms, the consistencies between the top 50 predicted miRNAs (obtained from LSGSP) and the results (confirmed from the updated HMDD, miR2Disease, and dbDEMC databases) reached 96% and 100%, respectively. Similarly, for isolated diseases (diseases not associated with any miRNAs), the consistencies between the top 50 predicted miRNAs (obtained from LSGSP) and the results (confirmed from the above-mentioned three databases) reached 98% and 100%, respectively. These results further indicate that LSGSP can effectively predict potential associations between miRNAs and diseases.
大量研究结果表明，微小RNA（miRNA）参与了许多重要的生物学过程；它们的突变和失调与疾病密切相关，因此，确定人类疾病与miRNA之间的关联是理解致病机制的关键。现有的用于识别miRNA与疾病关联的生物学实验方法通常既昂贵又耗时。因此，近年来，开发高效可靠的计算方法来识别与疾病相关的miRNA已成为生物学研究领域的一个重要课题。在本研究中，我们使用图的拉普拉斯分数和空间投影联合方法（LSGSP）开发了一种新型的miRNA与疾病关联预测模型。该模型整合了经实验验证的miRNA与疾病关联、疾病语义相似性分数、miRNA功能分数以及miRNA家族信息，构建了一个新的疾病相似性网络和miRNA相似性网络，然后通过计算图的拉普拉斯分数获得这些网络的全局相似性，在此基础上，结合miRNA与疾病布尔网络构建miRNA与疾病加权网络。最后，通过将miRNA空间和疾病空间投影到miRNA与疾病加权网络上获得miRNA与疾病分数。与其他几种先进方法相比，使用留一法交叉验证（LOOCV）在基准数据集、预测数据集和比较数据集上评估LSGSP的准确性，LSGSP分别以0.9221、0.9745和0.9194的高AUC值显示出优异的预测性能。此外，对于前列腺肿瘤和肺肿瘤，前50个预测的miRNA（从LSGSP获得）与结果（从更新的HMDD、miR2Disease和dbDEMC数据库确认）之间的一致性分别达到96%和100%。同样，对于孤立疾病（与任何miRNA均无关联的疾病），前50个预测的miRNA（从LSGSP获得）与结果（从上述三个数据库确认）之间的一致性分别达到98%和100%。这些结果进一步表明LSGSP可以有效地预测miRNA与疾病之间的潜在关联。
