Inferring Cell-type-specific Genes of Lung Cancer Based on Deep Learning
Jingyu Huang, Nitao Cheng, Chen Chen, Changsheng Li
Abstract
BACKGROUND: Lung cancer is cancer with the highest incidence in the world, and there is obvious heterogeneity within its tumor. The emergence of single-cell sequencing technology allows researchers to obtain cell-type-specific expression genes at the single-cell level, thereby obtaining information regarding the cell status and subpopulation distribution, as well as the communication behavior between cells. Many researchers have applied this technology to lung cancer research, but due to the shortcomings of insufficient sequencing depth, only a small part of the gene expression can be detected. Researchers can only roughly compare whether a few thousand genes are significant in different cell types. METHODS: To fully explore the expression of all genes in different cell types, we propose a method to predict cell-type-specific genes. This method infers cell-type-specific genes based on the expression levels of genes in different tissues and cells and gene interactions. At present, biological experiments have discovered a large number of cell-type-specific genes, providing a large number of available samples for the application of deep learning methods. RESULTS: Therefore, we fused Graph Convolutional Network (GCN) with Convolutional Neural Network( CNN) to build, model, and inferred cell-type-specific genes of lung cancer in 8 cell types. CONCLUSION: This method further analyzes and processes single-cell data and provides a new basis for research on heterogeneity in lung cancer tumor, microenvironment, invasion and metastasis, treatment response, drug resistance, etc.