Scoring Single-Sample Pathway Expression Level Using Graph Autoencoder

Single-sample pathway analysis (ssPA) is a bioinformatics technique used to assess the activity of biological pathways in individual samples, rather than relying on aggregated data from multiple samples. This approach allows for the detection of pathway activation or suppression in single samples, making it a valuable approach in research and clinical applications where individual variability is critical. In this paper, we propose a deep-learning method that scores individual pathway expression levels using a graph autoencoder. The proposed method provides insights into the biological processes and leverages the high dimensionality of gene expression data by setting the nodes in the neural network as biological pathways. The method also investigates unrevealed relationships among the biological pathways with graph convolutional layer while training the model. We evaluate the proposed method with four different single-sample pathway analysis studies: ssGSEA, GSVA, Pathifier, and kPCA. Since pathway analysis has no ground truth, we examine and perform various evaluation strategies to present robust results.