基于多源异构数据融合的miRNA与疾病高精度关联预测研究

Accumulating studies have shown that microRNAs (miRNAs) play critical roles in the development and progression of various human complex diseases. Unprecedented attention has been received to the highly accurate and efficient association prediction methods between miRNAs and diseases by using intelligent computing technology. Aiming at "The Study of Highly Accurate Prediction for Associations between miRNAs and Diseases Based on Multi-source Heterogeneous Data Fusion", This project is to adopt tensor decomposition to integrate multi-source heterogeneous biological data to an organic numerical representation scheme, and simulate human brain stratified structure to design the deep network model to automatically abstract feature vector with minimized reconstruction error. Finally, an acute and effective machine learning based prediction model will be constructed to identity the associations between miRNAs and diseases to provide theoretical and technical support for further development of biological experiments and precision medicine.

愈来愈多的研究表明microRNA（miRNA）在众多人类复杂疾病的发生发展过程中扮演着至关重要的作用。利用智能计算技术高效、精准地实现miRNA与疾病关联预测正在受到前所未有的关注与重视。本项目以“基于多源异构数据融合的miRNA与疾病高精度关联预测研究”为核心目标，拟提出一整套基于张量分解的miRNA和疾病相关多源异构生物数据的有机融合和数值化表示，模拟人脑分层结构设计基于深度学习的miRNA-疾病关联特征抽取模型自动提取具有最小化重构误差的抽象特征，设计基于机器学习的miRNA-疾病关联精准预测方法体系，为进一步开展定向生物实验和精准医疗提供理论和技术支撑。

识别 microRNA(miRNA)与疾病之间的关联性对于临床医疗具有极其重要的意义,因此构建 miRNA与疾病关联预测模型受到生物科学界越来越大的关注.对其研究将有助于更加深入的理解生命现象的本质，并对生命科学的发展起到强有力的推动作用。本项目以“基于多源异构数据融合的miRNA与疾病高精度关联预测研究”为目标， 融合 miRNA 与疾病相关的多源异构生物数据，挖掘多源异构数据中 miRNA-疾病关联的数值化本征表达，设计并实现了多种基于机器学习和深度学习的预测模型，实现了miRNA-疾病关联的高效预测。实验结果表明这些模型可以作为指导未来潜在miRNA生物标记物的生物实验验证的有用工具。本项目的研究成果对于加深对生命的认识，促进对疾病潜在关联 miRNA 的辨识，有效降低定向生物实验研究的时间和成本，为推动人类复杂疾病的个体化精准医疗和生物制药提供科学依据和理论支撑。