School of Computer Science and Technology, China University of Mining and Technology, Xuzhou 221116, China.
Xinjiang Technical Institutes of Physics and Chemistry, Chinese Academy of Sciences, Urumqi 830011, China.
Mol Ther. 2021 Apr 7;29(4):1501-1511. doi: 10.1016/j.ymthe.2021.01.003. Epub 2021 Jan 9.
It is reported that microRNAs (miRNAs) play an important role in various human diseases. However, the mechanisms of miRNA in these diseases have not been fully understood. Therefore, detecting potential miRNA-disease associations has far-reaching significance for pathological development and the diagnosis and treatment of complex diseases. In this study, we propose a novel diffusion-based computational method, DF-MDA, for predicting miRNA-disease association based on the assumption that molecules are related to each other in human physiological processes. Specifically, we first construct a heterogeneous network by integrating various known associations among miRNAs, diseases, proteins, long non-coding RNAs (lncRNAs), and drugs. Then, more representative features are extracted through a diffusion-based machine-learning method. Finally, the Random Forest classifier is adopted to classify miRNA-disease associations. In the 5-fold cross-validation experiment, the proposed model obtained the average area under the curve (AUC) of 0.9321 on the HMDD v3.0 dataset. To further verify the prediction performance of the proposed model, DF-MDA was applied in three significant human diseases, including lymphoma, lung neoplasms, and colon neoplasms. As a result, 47, 46, and 47 out of top 50 predictions were validated by independent databases. These experimental results demonstrated that DF-MDA is a reliable and efficient method for predicting potential miRNA-disease associations.
据报道，微小 RNA（miRNA）在各种人类疾病中发挥着重要作用。然而，miRNA 在这些疾病中的机制尚未完全阐明。因此，检测潜在的 miRNA-疾病关联对病理发展以及复杂疾病的诊断和治疗具有深远的意义。在这项研究中，我们提出了一种新的基于扩散的计算方法 DF-MDA，用于基于假设分子在人类生理过程中相互关联的假设来预测 miRNA-疾病关联。具体来说，我们首先通过整合 miRNA、疾病、蛋白质、长非编码 RNA（lncRNA）和药物之间的各种已知关联来构建一个异构网络。然后，通过基于扩散的机器学习方法提取更具代表性的特征。最后，采用随机森林分类器对 miRNA-疾病关联进行分类。在 5 折交叉验证实验中，所提出的模型在 HMDD v3.0 数据集上获得了 0.9321 的平均曲线下面积（AUC）。为了进一步验证所提出模型的预测性能，DF-MDA 应用于三种重要的人类疾病，包括淋巴瘤、肺癌和结肠癌。结果，前 50 个预测中有 47、46 和 47 个得到了独立数据库的验证。这些实验结果表明，DF-MDA 是一种可靠且高效的预测潜在 miRNA-疾病关联的方法。
