CellVGAE uses the connectivity between cells (such as k-nearest neighbour graphs) with gene expression values as node features to learn high-quality cell representations in a lower-dimensional space