MetaGXplore: Integrating Multi-Omics Data with Graph Convolutional Networks for Pan-cancer Patient Metastasis Identification

Abstract

ObjectiveTumor metastasis, the spread of cancer cells from the primary tumor to distant anatomical sites, poses a significant challenge in clinical prognosis, impairing treatment efficacy and reducing patient survival time. Current methods for predicting and diagnosing tumor metastasis rely heavily on extensive examinations and subjective clinical judgments, introducing considerable uncertainty and inefficiency. Accurate prediction of tumor metastasis likelihood is crucial for guiding effective clinical interventions and extending patient survival. Additionally, identifying key genes that influence metastasis probability could provide valuable data for detecting tumor-specific biomarkers indicative of metastasis likelihood.MethodsMulti-omics datasets, which encompass diverse genomic information, offer a comprehensive understanding of cancer biology. Leveraging deep learning algorithms enhances the accuracy and predictive power in forecasting cancer metastasis, given their superior performance in complex prediction tasks. We developed a deep learning-based framework, MetaGXplore, which represents the first attempt to predict metastasis probability using pan-cancer multi-omic datasets and to explain the pivotal genes involved in the process.ConclusionsWe trained the classification model using a K-fold cross-validation procedure and interpreted the predicted results with GNNExplainer. The effectiveness of MetaGXplore was demonstrated by evaluating the model, graph structure contribution, and multi-omics data. Finally, enrichment analysis of differentially expressed genes and pathways was conducted to explore the biological functions of major differential genes and the relationships between key pathways.