Department of Molecular Microbiology and Immunology, Keck School of Medicine, University of Southern California, Los Angeles, USA.
Sci Rep. 2023 Jul 26;13(1):12093. doi: 10.1038/s41598-023-39282-4.
Single cell RNA sequencing has a central role in immune profiling, identifying specific immune cells as disease markers and suggesting therapeutic target genes of immune cells. Immune cell-type annotation from single cell transcriptomics is in high demand for dissecting complex immune signatures from multicellular blood and organ samples. However, accurate cell type assignment from single-cell RNA sequencing data alone is complicated by a high level of gene expression heterogeneity. Many computational methods have been developed to respond to this challenge, but immune cell annotation accuracy is not highly desirable. We present ImmunIC, a simple and robust tool for immune cell identification and classification by combining marker genes with a machine learning method. With over two million immune cells and half-million non-immune cells from 66 single cell RNA sequencing studies, ImmunIC shows 98% accuracy in the identification of immune cells. ImmunIC outperforms existing immune cell classifiers, categorizing into ten immune cell types with 92% accuracy. We determine peripheral blood mononuclear cell compositions of severe COVID-19 cases and healthy controls using previously published single cell transcriptomic data, permitting the identification of immune cell-type specific differential pathways. Our publicly available tool can maximize the utility of single cell RNA profiling by functioning as a stand-alone bioinformatic cell sorter, advancing cell-type specific immune profiling for the discovery of disease-specific immune signatures and therapeutic targets.
单细胞 RNA 测序在免疫分析中具有核心作用，可将特定免疫细胞鉴定为疾病标志物，并提示免疫细胞的治疗靶基因。单细胞转录组学中的免疫细胞类型注释对于从多细胞血液和器官样本中解析复杂的免疫特征非常重要。然而，仅从单细胞 RNA 测序数据进行准确的细胞类型分配，由于基因表达异质性水平较高而变得复杂。已经开发出许多计算方法来应对这一挑战，但免疫细胞注释的准确性并不高。我们提出了 ImmunIC，这是一种通过结合标记基因和机器学习方法来识别和分类免疫细胞的简单而强大的工具。通过 66 项单细胞 RNA 测序研究中的超过 200 万个免疫细胞和 50 万个非免疫细胞，ImmunIC 在识别免疫细胞方面的准确率达到 98%。ImmunIC 优于现有的免疫细胞分类器，可将其分类为十种免疫细胞类型，准确率为 92%。我们使用先前发表的单细胞转录组学数据来确定严重 COVID-19 病例和健康对照的外周血单核细胞组成，从而能够识别免疫细胞类型特异性差异途径。我们的公开工具可以通过作为独立的生物信息学细胞分选器发挥作用，最大程度地提高单细胞 RNA 分析的效用，从而推进针对疾病特异性免疫特征和治疗靶标的细胞类型特异性免疫分析。
