Single-cell data analysis is widely used to reveal the heterogeneity and dynamics of tissues, organisms, and complex diseases. Deep learning provides a great tool to uncover intricate biological patterns and relationships underlying large-scale, noisy single-cell data. We introduced scGNN as a hypothesis-free graph neural network framework for single-cell RNA-Seq analyses. It integrates three iterative multi-modal autoencoders and outperforms existing tools for gene imputation and cell clustering on four benchmark single-cell RNA-Seq datasets. We further developed RESEPT, a deep-learning framework for characterizing and visualizing tissue architecture from spatially resolved transcriptomics by reconstructing and segmenting a transcriptome mapped RGB image. RESEPT can identify the tissue architecture, and represent corresponding marker genes and biological functions accurately. We also developed DeepMAPS for biological network inference from single-cell multi-omics data. By building a heterogeneous graph containing both cell and gene nodes, DeepMAPS identifies the joint embedding of all the nodes simultaneously and enables the inference of cell-type-specific biological networks. These tools provide critical insights into the underlying mechanisms driving the complex tissue heterogeneities in development and diseases.