基于双层生物模块网络的致病基因预测

Disease Genes Prediction Based on Two-layer Biological Module Network

为提高致病基因预测的准确性,提出了一种结合多种生物组合数据的生物分子网络构建方法。利用蛋白质相互作用网络、蛋白质复合物和代谢通路构建双层生物模块网络,并设计一种新型异构图神经网络。实验结果表明,与基于蛋白质相互作用组学数据构建生物分子网络的经典方法相比,本方法的AUC和F1分数提高了约1.5%和3.7%。

In order to improve the accuracy of disease genes prediction,a method of constructing biological network combining multiple biological data was proposed.A two-layer biological module network was constructed by protein interaction,protein complexes,metabolic pathways and designed a novel graph neural network.Experimental results show that the AUC and F1 scores of this method are improved by about 1.5%and 3.7%,compared with the classical methods of constructing biological networks based on protein interaction data.