Decoding the gene subnetworks prioritized in prediction with GLRP in Alzheimer’s Disease

Abstract

AbstractBackgroundAlzheimer’s disease is the leading cause of brain dementia, along with which substantial failure of organs and mental issues arise. The abundance of AD related data in the current decade has allowed for much advancement in the field using modern machine learning and deep learning techniques to decrypt pathology. Though diagnostic tools have been modeled over such large expanses of data, the black box problem of decoding the significant biological components contributing to the prediction has been overlooked in the research field.MethodGLRP is a method to decode the final predictions of a deep learning method. Since Graph‐CNN are best suited for using the gene expression data modeled over an established whole human PPI network. The usage of GLRP allows us to build the relevance that we need to interpret how the genes contribute to the prediction based on proven PPI network topology as the basis. From the relevance’s obtained of correctly predicted ROSMAP AD subjects and controls, important genes were extracted to be analyzed for functional annotations and pathway analysis with HPRD PPI.ResultThe graph‐CNN model was able to predict between AD subjects and controls with an AUC of 76.58% and test accuracy of 79.167% while the highest accuracy of machine learning models capped at 75% with an AUC of 72.95% from SVM as seen in Table 1. Having found a threshold of 550 genes to have the highest connected component in graph before noisy genes are added to the network as seen in Fig 1, the average relevance of AD subjects and controls provided us with 214 common genes with 181 genes in the largest connected component as seen in Fig 2.ConclusionAlong with tackling the black box problem of the deep learning model using GLRP which has allowed us to gather the important genes from the network view at a higher efficiency than the machine learning models. The top 214 common genes and the network relating to the negative regulation of organelle organization, hypoxia and oxygen level response, and cellular response to tumor necrosis factors could have potential impact on the pathology of AD and its progression.