Identification of cell-type-specific genes in multimodal single-cell data using deep neural network algorithm

• Published in: Computers in biology and medicine Vol. 166; p. 107498
• Main Authors: Qian, Weiye, Yang, Zhiyuan
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.11.2023; Elsevier Limited
• Subjects: Algorithms; Artificial neural networks; Bioinformatics; Bone marrow; Bone Marrow Cells - cytology; Bone Marrow Cells - metabolism; Cell differentiation; Datasets; Deep Learning; Differentiation (biology); Epitopes; Gene expression; Gene Expression Profiling - methods; Gene sequencing; Genes; Heterogeneity; Humans; Internal Medicine; Machine learning; Multimodal single-cell study; Neural networks; Neural Networks, Computer; Neural stem cells; Neutrophils; Optimization techniques; Other; Proteins; Ribonucleic acid; RNA; RNA-protein interactions; Sequence Analysis, RNA - methods; Single-Cell Analysis - methods; Stem cells; Student's t-test; Transcriptome - genetics; Transcriptomes; Deep learning; Bioinformatics; Multimodal single-cell study
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2023.107498

The emergence of single-cell RNA sequencing (scRNA-seq) technology makes it possible to measure DNA, RNA, and protein in a single cell. Cellular Indexing of Transcriptomes and Epitopes by sequencing (CITE-seq) is a powerful multimodal single-cell research innovation, allowing researchers to capture RNA and surface protein expression on the same cells. Currently, identification of cell-type-specific genes in CITE-seq data is still challenging. In this study, we obtained a set of CITE-seq datasets from Kaggle database, which included the sequencing dataset of seven cell types during bone marrow stem cell differentiation. We used Student's t-test to analyze these transcription RNAs and pick out 133 significantly differentially expressed genes (DEGs) among all cell types. Functional enrichment revealed that these DEGs were strongly associated with blood-related diseases, providing important insights into the cellular heterogeneity within bone marrow stem cells. The relation between RNA and protein levels was performed by deep neural network (DNN) model and achieved a high prediction score of 0.867. Based on their coefficients in the DNN model, three genes (LGALS1, CENPV, TRIM24) were identified as cell-type-specific genes in erythrocyte progenitor. Our works provide a novel perspective regarding the differentiation of stem cells in the bone marrow and provide valuable insights for further research in this field. •The multimodal CITE-seq datasets were obtained and analyzed.•A novel quantitative model of RNA and protein was built by deep neural network algorithm.•18 cell-type-specific genes were identified in bone marrow stem cells.