AutoGGN: A Gene Graph Network AutoML Tool for Multi-Omics Research

Abstract

Omics data identifies biological characteristics from genetic to phenotypic levels during the life span. Molecular interaction networks have a fundamental impact on life activities. Integrating omics data and molecular interaction networks will help researchers delve into comprehensive information underlying the data. Here, we proposed a new multimodal method called AutoGGN to aggregate multi-omics data and molecular interaction networks based on graph convolutional neural networks. We evaluated AutoGGN using two different tasks: cancer type classification and single-cell stage classification. On both tasks, AutoGGN showed better performance compared to other methods, the trend is relevant to the ability of utilizing much more information from biological data. The phenomenon indicated AutoGGN has the potential to incorporate valuable information from molecular interaction networks and multi-omics data effectively. Furthermore, in order to provide a better understanding of the mechanism of prediction results, we assessed the explanation using SHAP module and identified the key genes contributing to the prediction of classification, which will provide insights for the downstream design of biological experiments.