GCNG: graph convolutional networks for inferring gene interaction from spatial transcriptomics data

Abstract

Most methods for inferring gene-gene interactions from expression data focus on intracellular interactions. The availability of high-throughput spatial expression data opens the door to methods that can infer such interactions both within and between cells. To achieve this, we developed Graph Convolutional Neural networks for Genes (GCNG). GCNG encodes the spatial information as a graph and combines it with expression data using supervised training. GCNG improves upon prior methods used to analyze spatial transcriptomics data and can propose novel pairs of extracellular interacting genes. The output of GCNG can also be used for downstream analysis including functional gene assignment.Supporting website with software and data: https://github.com/xiaoyeye/GCNG.

Computational challenges and opportunities in spatially resolved transcriptomic data analysis

“…With single-cell sequencing data, cell–cell communication inference has relied on identifying coordinated expression of known ligand–receptor pairs 8 . Computational methods that leverage the added spatial information from spatially resolved transcriptomic data using graph convolutional neural networks 9 , optimal transport approaches 10 , and spatial cross-correlation analysis 6 can narrow down candidates to ligand–receptor pairs that are spatially colocalized, potentially indicative of autocrine or paracrine signaling. Furthermore, spatially resolved transcriptomic data with co-registered imaging data present additional sources of heterogeneity, such as morphological variability, which can be used for clustering, as differences in morphology can be a proxy for differences in cell states or other functional phenotypes such as cell cycle position, transformation, or invasiveness.…”

Integrating single-cell and spatial transcriptomics to elucidate intercellular tissue dynamics

“…Such deep learning approaches, if successfully applied on large series of well-characterized data sets, may complement integrated scRNA-seq and spatial transcriptomics data sets by capturing information on cell types and gene expression via conventional histology. In addition to improving deconvolution and mapping algorithms, one needed focus centres on developing additional deep learning models [138][139][140][141] to help disentangle which features of a given spatial transcriptome are most biologically relevant.…”

A Biologically Interpretable Graph Convolutional Network to Link Genetic Risk Pathways and Neuroimaging Markers of Disease

“…Graph convolutional networks (GCNs) [30] provide a natural means to incorporate topological information and data geometry in the form of connections between the nodes. GCNs have gained immense popularity in protein structure prediction [14], drug discovery [45], and gene-gene interactions [53]. Beyond the architectural design, graph attention layers learn a set of "edge importance" weights, which enable us to track the information flow through the graph [48].…”