Graph neural network based AD diagnosis prediction.
Current genotype-to-phenotype models, such as polygenic risk scores, only account for linear relationships, limiting the complexity of the diseases that can be properly characterized. In addition, previous biological information, such as protein-protein interactions (PPIs), is not generally used and models ignore epistatic interactions. However, interactions at the protein level can have profound implications in understanding the genetic etiology of diseases and, in turn, for drug development. Thus, here we propose an approach for phenotype prediction based on graph neural networks (GNNs) that naturally incorporates existing PPI information into the model. As a result, our approach can naturally discover relevant epistatic interactions. We assess the potential of this approach using simulations and comparing it to linear and other non-linear approaches. We also study the performance of the proposed GNN-based methods in predicting Alzheimer's disease, one of the most complex neurodegenerative diseases. We show in both cases that GNN-based approaches outperform the state of the art. In addition, we show that, we can discover critical interactions in the disease. Our findings highlight the potential of GNNs in predicting phenotypes and discovering the underlying mechanisms of complex diseases.