针对疾病检测方法中忽略微生物潜在关联问题,提出一种基于微生物转移网络的新型图神经网络算法(SAGE-O-N)用于疾病检测.算法利用微生物测序信息的相似性特征构建微生物转移网络,使用图结构数据挖掘算法发掘相似性与表型关联.实验结果表明,与传统机器学习方法相比,SAGE-O-N 在单种疾病上的受试者工作特征曲线面积(Area Under Curve,AUC)提高了约 2%,在并发症数据集中AUC提高约 4%.
Disease Detection Algorithms Based on Microbial Transfer Network
A novel graph neural network algorithm(SAGE-O-N)based on microbial transfer network was proposed for disease detection,which addressed the problem of ignoring potential microbial associations in disease detection methods.The similarity features of microbial sequencing information was utilized to con-struct a microbial transfer network,and a graph-structured data mining algorithm was used to uncover similarity and phenotypic associations.Experimental results show that SAGE-O-N improves the area un-der curve(AUC)of subjects'working characteristics on a single disease by about 2%,and the AUC in the comorbidity dataset by about 4%,compared with traditional machine learning methods.
graph neural networkdisease detectionmicrobiota sequencingmicrobial transfer network
国家科技期刊平台
NSTL
万方数据
维普
