Skip to main content
NCBI home page
Clinical and Translational Medicine logoLink to Clinical and Translational Medicine
letter
. 2022 Dec 10;12(12):e1132. doi: 10.1002/ctm2.1132

Evaluation of pulmonary single‐cell identity specificity in scRNA‐seq analysis

Xuanqi Liu 1,2,3, Guang Xu 4, Chengshui Chen 5,6,✉, Yuanlin Song 1,2,3,✉, Wei Wang 4,✉, Xiangdong Wang 1,2,3,✉
PMCID: PMC9736794  PMID: 36495131

Dear Editor,

The single cell RNA sequencing (scRNA‐seq) technology provides new insights into understanding of single‐cell transcriptomic atlas and intercellular communication. 1 scRNA‐seq is used to characterize the considerable heterogeneity and complexity of cell type, to uncover the cell fate and context, critical molecular features and progression trajectories, as well as to explore potential pathogenesis and individualized therapeutic targets. 2 , 3 , 4 One of major challenges is to identify of biology‐specific biomarkers as the cell identity for cell phenotypes and functional sub‐types, although bioinformatic and fstatistical methods for scRNA‐seq are developed and improved rapidly. 5 More and more cell type/subtypes are identified in response to various stimulus, external challenge and even pathological conditions. The correctness and specificity of cell‐specific transcriptomic profiles based on scRNA‐seq are highly dependent upon the accuracy of cell type identity and cellular annotation. The process of mapping single cell atlas is based on selected marker genes. Intricated cell types and subtypes are defined by the provided cell annotations and further validated using cell identity marker gene panels (ciMGPs), 6 so the specificity of ciMGPs is critical to construct the single‐cell profile.

The current studies aim at evaluating the specificity of ciMGPs for various pulmonary single‐cell identities and to define disease‐specific alterations of single cell populations labelled with ciMGPs. We screened and selected 57 ciMGPs from previous studies 6 , 7 , 8 and validated the identities of ciMGPs‐based cell types/subtypes in lung tissues from healthy subjects or patients with idiopathic pulmonary fibrosis (IPF), chronic obstructive pulmonary disease (COPD), systemic sclerosis (SSC), lung adenocarcinoma (LUAD), large cell cancer (LCC) or para‐cancer tissues as pair‐controls, as detailed in Supplemental Materials. Of those cells, immune cells resident in the lung tissue included nine subtypes of lymphoid cells and 15 subtypes of myeloid cells, 8 and lung parenchymal cells had 15 subtypes of epithelia, nine of endothelia and nine of stromal cells, as presented in Figures S1–S48 and details in Tables S6–S11. We comprehensively assessed and quantified the specificity of each ciMGP representing 57‐specific cell subtypes among lung diseases. We firstly developed the criteria and schematic diagram to determine the specificity and accuracy of lung cell ciMGPs, as explained in Supplemental Method. The scRNA‐seq data for evaluation were collected from various databases (Tables S1–S3). We defined the overlap expression rate (OER) of ciMGPs in a cell subtype was less than 5%, as the cell‐specific marker panel with high specificity, when compared with the expression in other cell types/subtypes, between 5% and 10% as the ‘cell‐associated marker panel’ with moderate specificity, or more than 10% as ‘cell‐reference marker panel’ with low specificity, as explained in Supplemental Method and Figure 5. We dedicate special attention to the alteration of ciMGP's specificity in illness states, which presents profound insight into the clinical promotion and popularization.

FIGURE 5.

FIGURE 5

Open in a new tab

The flowchart showing the process of scRNA‐seq data collection (A), sample integration and quality control (B), data scalation (C), clustering analysis (D) and evaluation of marker gene panel specificity (E) among cells and diseases, based on the value of overlap expression rate (OER). OER is calculated from the overlapped marker cell subtype number/total cell subtype number *100%, and the indicator of the proportion ratio of MGP specificity in certain cell types/subtypes.

We established the transcriptomic profiles of pulmonary single cells based on unified manifold approximation and projection (UMAP), reflecting the abundance and distribution of cell types and subtypes (Figure 1). The specificity of ciMGPs was evaluated in different cell types, subtypes, locations and diseases. The OER values of B cell (Figure 2A) and adventitial fibroblast (Figure 2B) were less than 5% in normal lung tissue or various lung diseases.

FIGURE 1.

FIGURE 1

Open in a new tab

The uniform manifold approximation and projection (UMAP) representation of the lung cell landscape. Unsupervised analysis of 57 pulmonary single‐cells including 15 subtypes of epithelial cell, nine of endothelial cells, nine of stromal cells and 24 of immune cells harvested from lung tissues of healthy controls (normal) or patients with chronic obstructive pulmonary disease (COPD), idiopathic pulmonary fibrosis (IPF), lung adenocarcinoma (LUAD), large cell lung cancer (LCC), para‐tumor tissue (Para) and systemic sclerosis (SSC) and their corresponding characterization.

FIGURE 2.

FIGURE 2

Open in a new tab

Cell‐specific panel in lung diseases. The box plot indicated the average mRNA expression of the panel in B cell (A; CD79A, MS4A1, CD19), the panel in adventitial fibroblast (B; SERPINF1, PI16, FBLN1, SCRAR5) and the panel in AT1 (C; HOPX, PDPN, CLIC5, AGER, CLDN18, EMP2) in 57 cell types from normal lung samples and six lung diseases (samples of para‐cancer, LCC, LUAD, IPF, COPD, SSC). The RED bars represent panels of B cell, AT1 and adventitial fibroblast, as the cell‐specific panel with <5% overlap expression rate as compared to other cell subtypes. The detailed calculation procedure of overlap expression rate can be seen in the Supplementary Methods.

The AT1 ciMGPs showed highly specific in most samples except for LCC, as compared with the remaining 56 cell subtypes (Figure 2C). Of 15 subtypes of lung epithelial cells, the ciMGPs specificity of ciliated cell was the highest in eight lung tissues (Table 1). The basal epithelia failed to be detected in LUAD samples based on the provided panel, and the efficacy of panel in goblet epithelia was weakened in LCC samples, according to the criteria proposed (Table 1). The specificity of ciMGPs representing pericyte cells and artery endothelial cells were comparatively decreased in the states of LCC and SSC (Table 1, Figures S4 and S35). However, ciMGPs mRNA expression of bronchial vessel 2 and pulmonary inocyte was hardly detected in all kinds of cells and lung diseases in our research (Table 1, Figures S8 and S24). The representatives of cell‐specific panels in the process of cell annotation were summarized and presented in Tables S4 and S5.

TABLE 1.

‘Overlap expression cell subset rate’ of cell subset‐specific, cell subset‐associated, cell subset‐reference mark gene panels of human lung tissues harvested from patients with lung adenocarcinoma (LUAD), large cell cancer (LCC), idiopathic pulmonary fibrosis (IPF), chronic obstructive pulmonary disease (COPD), and systemic sclerosis (SSC) total, normal (norm), and para‐cancer human lung tissues

Cell type Total Normal Para‐cancer LCC LUAD IPF COPD SSC Cell‐specific (<5%) Cell‐associated (5%–10%) Cell‐reference (>10%)
Adventitial fibroblast 0 0 0 3.57 3.57 1.79 0 1.79 8 0 0
Airway smooth muscle cell 1.79 3.57 0 1.79 0 1.79 1.79 1.79 8 0 0
Alveolar epithelial type 1 0 0 0 41.07 5.36 1.79 0 0 6 1 1
Alveolar epithelial type 2 1.79 1.79 1.79 5.36 1.79 1.79 1.79 1.79 7 1 0
Alveolar fibroblast 1.79 3.57 0 17.86 1.79 1.79 0 1.79 8 0 0
Artery endothelia 0 0 0 3.57 0 0 0 1.79 8 0 0
B cell 0 0 0 1.79 1.79 0 0 0 8 0 0
Basal epithelia 1.79 3.57 0 0 ND 3.57 3.57 3.57 7 0 0
Basophil/mast 1 1.79 3.57 0 ND ND 3.57 3.57 3.57 6 0 0
Basophil/mast 2 1.79 1.79 1.79 1.79 1.79 1.79 1.79 1.79 8 0 0
Bronchial vessel endothelia 1 33.93 26.79 ND 3.57 25 5.36 30.36 16.07 1 1 5
Bronchial vessel endothelia 2 ND ND ND ND ND ND ND ND
Capillary endothelia 8.93 7.14 5.36 14.29 14.29 23.21 7.14 3.57 1 4 3
Capillary aerocyte 0 1.79 1.79 0 0 0 1.79 1.79 8 0 0
Capillary Intermediate endothelia 1 8.93 5.36 3.57 ND 5.36 7.14 5.36 12.5 1 5 1
Capillary intermediate endothelia 2 30.36 23.21 28.57 21.43 26.79 37.5 23.21 30.36 0 0 8
CD4+ memory/effector T cell 28.57 35.71 21.43 21.43 12.5 41.07 85.71 25 0 0 8
CD4+ naive T cell 73.21 76.79 60.71 14.29 14.29 91.07 76.79 62.5 0 0 8
CD8+ memory/effector T cell 10.71 10.71 10.71 10.71 10.71 10.71 10.71 8.93 0 1 7
CD8+ Naive T cell 8.93 5.36 8.93 8.93 8.93 7.14 7.14 10.71 0 7 1
Ciliated epithelia 3.57 1.79 0 0 0 3.57 3.57 3.57 8 0 0
Classical monocyte 5.36 3.57 1.79 3.57 3.57 5.36 1.79 1.79 6 2 0
Club epithelia 10.71 5.36 1.79 67.86 12.5 7.14 5.36 7.14 1 4 3
Differentiating basal epithelia 14.29 19.64 ND 17.86 14.29 17.86 23.21 21.43 0 0 7
EREG+ dendritic 12.5 14.29 10.71 10.71 7.14 14.29 10.71 10.71 0 1 7
Fibromyocyte 5.36 7.14 ND 5.36 5.36 5.36 12.5 7.14 0 6 1
Goblet epithelia 3.57 0 1.79 5.36 3.57 3.57 1.79 3.57 7 1 0
IGSF21+ dendritic 16.07 16.07 10.71 10.71 14.29 16.07 14.29 16.07 0 0 8
Intermediate monocyte 25 26.79 28.57 14.29 16.07 21.43 25 19.64 0 0 8
Ionocyte ND ND ND ND ND ND ND ND
Lipofibroblast 78.57 ND ND ND 89.29 96.43 ND 42.86 0 0 4
Lymphatic endothelia 0 0 0 12.5 0 0 0 0 7 1 0
Macrophage 5.36 7.14 0 14.29 5.36 7.14 3.57 8.93 2 5 1
Mesothelial cell 1.79 0 1.79 21.43 12.5 0 0 ND 5 0 2
Mucous epithelia 3.57 3.57 3.57 5.36 3.57 3.57 1.79 3.57 7 1 0
Myeloid dendritic type 1 8.93 8.93 10.71 8.93 7.14 8.93 10.71 8.93 6 2 0
Myeloid dendritic type 2 35.71 23.21 21.43 33.93 30.36 37.5 30.36 16.07 0 0 8
Myofibroblast 7.14 8.93 3.57 7.14 8.93 10.71 12.5 7.14 1 5 2
Natural killer cell 35.71 35.71 32.14 57.14 33.93 35.71 64.29 26.79 0 0 8
Natural killer T cell 44.64 37.5 35.71 32.14 35.71 44.64 39.29 41.07 0 0 8
Neuroendocrine epithelia 0 0 ND ND ND ND 0 ND 3 0 0
Non‐classical monocyte 94.64 85.71 67.86 23.21 94.64 96.43 42.86 91.07 0 0 1
OLR1+ classical monocyte 21.43 21.43 14.29 5.36 16.07 23.21 17.86 21.43 0 1 7
Pericyte cell 0 0 0 1.79 0 0 0 1.79 8 0 0
0Plasma cell ND ND ND ND ND ND ND ND
Plasmacytoid dendritic ND ND 12.5 ND ND ND ND ND 0 0 1
Platelet/Megakaryocyte 26.79 28.57 ND ND 19.64 25 ND 26.79 0 0 5
Proliferating basal epithelia ND ND ND ND ND ND ND ND
Proliferating macrophage 10.71 8.93 ND ND ND 8.93 ND 16.07 0 2 2
Proliferating NK/T cell 46.43 58.93 50 37.5 41.07 44.67 53.57 ND 0 0 7
Proximal basal epithelia 1.79 5.36 ND ND ND 3.57 3.57 3.57 4 1 0
Proximal ciliated epithelia 1.79 1.79 ND ND ND 1.79 3.57 3.57 5 0 0
Serous epithelia ND ND 0 3.57 7.14 ND 7.14 ND 2 2 0
Signaling alveolar epithelial type 2 1.79 1.79 1.79 0 48.21 7.14 1.79 1.79 6 1 1
TREM2+ dendritic 14.29 23.21 25 14.29 16.07 17.86 25 17.86 0 0 8
Vascular smooth muscle cell 3.57 5.36 3.57 1.79 3.57 3.57 1.79 3.57 7 1 0
Vein endothelia 21.43 8.93 1.79 35.71 16.07 17.86 17.86 14.29 1 1 06
Cell‐specific (<5%) 22 21 26 14 15 20 24 22
Cell‐associated (5%–10%) 8 12 2 8 9 10 5 6
Cell‐reference (>10%) 21 17 16 22 22 20 20 20
Open in a new tab

The specificity and importance of ciMGPs can be dynamically changed on basis of disease nature and severity. The specificity of ciMGPs appeared in myeloid cells (Figure 3C) was higher than lymphoid cells (Figures 2A and 4B). The quality of ciMGPs of signaling AT2, CD8 naïve T cell and classic monocyte were considered as ‘cell‐associated panels’ in the normal lung tissue, while became the cluster of ‘cell‐specific panels’ in lung diseases and varied among lung diseases (Figure 3). The signaling AT2 panel clearly up‐expressed in AT2 from normal and para‐cancer lung tissues while became unclear in LUAD and IPF (Figure 3A). Compared with tissue‐resident cells, the ciMGP of immune cells, including CD4 memory effector T, CD4 naïve T, CD8 memory effector T, NK or NKT, showed relatively lower specificity, was more difficult to be annotated, meanwhile, and highly expressed in 2–3 other cell subtypes (Figures S12, S13 and S32, Table 1). It might be attributed to the relatively conserved function of structural cells in lungs, while immune cells exist in intermediate and functional states with continuously dynamical remodeling. Immune cells can be activated by external stimuli, to perform the primary force of host defense in lung, including the process of rapid recruitment and migration. 9 Our data demonstrated that the ciMGPs specificity of capillary intermediate endothelia 2, natural killer and lipofibroblast was low in the majority of lung samples (Figure 4A–C), as the cell‐reference ciMGPs with high OERs. The specificity of ciMGPs has the value for deeply understanding the heterogeneity among various lung diseases and pathological states. Several ciMGPs with tissue‐specific pattern of expression have the potential of clinical implications.

FIGURE 3.

FIGURE 3

Open in a new tab

Cell‐associated panel in lung diseases. The box red plot represents the panel gene expression of signaling_AT2 (A; CP, SFTPC), CD8 naïve T cell (B; CD3E, GZMH, GZMB) and classic monocyte (C; LGALS2, CD14, NRG1, S100A8, S100A9, S100A12) harvested from normal lung samples and six lung diseases (samples of para‐cancer, LCC, LUAD, IPF, COPD, SSC). The panels of signaling_AT2, CD8 naïve T cell and classic monocyte were presented as examples of cell‐associated panel with 5%–10% overlap expression rate. The detailed calculation procedure of overlap expression rate can be seen in the supplemental methods.

FIGURE 4.

FIGURE 4

Open in a new tab

Cell‐reference panel in lung diseases. The box red plot indicated the panel gene expression in capillary intermediate endothelia 2 (A; IL7R, SLC6A4, FCN3), natural killer cell (B; FCER1G, TROBP, GZMB, CHST2, HOPX) and lipofibroblast (C; PLIN2, APOE) from normal lung samples and six lung diseases (samples of para‐cancer, LCC, LUAD, IPF, COPD, SSC). The panels of capillary intermediate endothelia 2, natural killer cell and lipofibroblast were presented as examples of cell‐reference panel with more than 10% overlap expression rate. The detailed calculation procedure of overlap expression rate can be seen in the Supplemental Methods.

The ciMGPs of tissue‐resident cells and immune cells showed obvious differences in disease specificity, especially the subtypes of serous and CD4 naïve T cells, of which the specificity of ciMGPs was significantly higher in LUAD and LCC tissues (Table 1). The ciMGPs of AT1, Club or mesothelial cells showed comparatively low specificity in LCC samples (Table 1, Figures S17 and S27), while the ciMGPs of CD4 naïve T and non‐classic monocytes were higher in LCC (Table 1). Some ciMGPs expressed in multiple cell types, subtypes and diseases, for example, capillary intermediate endothelia 2 (Figure 4A), natural killer cells (Figure 4B) and lipofibroblast (Figure 4C).

It is critical to evaluate the specificity of ciMGPs in normal tissues to precisely define the representatives of cells and set the referenced baseline, in pathological disease tissues to check the abnormal values, and in development‐related cell subtypes to clarify the OERs. The cells with similar developmental‐lineage often share the common canonical molecular markers and resemble in their expression patterns, which makes it difficult to differ between AT1 and AT2, basophil/mast1 and 2, vascular and airway smooth muscle cells, proximal basal and basal cells, or myeloid dendritic type I and II. We found the variation range of ciMGPs among multiple cell subtypes and even types, for example, ciMGPs signature of dendritic cells also highly expressed in alveolar epithelial cells and basophil/mast cells (Figures S2, S5 and S6). It implied that they might share the common underlying gene expression pattern and close molecular interaction or crosstalk among diverse types of cells. The novelty of the present study is to comprehensively define and evaluate ciMGP specificity of pulmonary single‐cells and proposed the three categories using the OER values, to determine the difference of ciMGP specificity among multiple pathological conditions and to provide new alternatives for the quality control in scRNA‐sq data analysis and clinical application, as proposed previously. 10 However, limited scRNAseq data set might impact the extrapolation of the conclusion. The specificity of distinct stages, surgery procedures, lesioned sites of lung diseases and even data derived from different sequencing methods needs to be further validated.

In conclusion, for the first time, we developed the criteria to evaluate the ciMGPs specificity of lung cell types/subtypes from various lung diseases and characterized three categories of cell‐specific, cell‐associated, and cell‐reference ciMGPs on basis of scRNA‐seq. The ciMGPs specificity varied among cell types and subtypes, disease natures and stages, as well as responses to therapies as the part of quality control in scRNA‐seq analysis, although the evaluation and criteria of ciMGPs need to be further improved and optimized. Thus, we believe that the precise evaluation of ciMGPs specificity is considerably important in bioinformatic analysis, single cell categories, data interpretations and accurate conclusion.

CONFLICT OF INTEREST

The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

Supporting information

Supporting Information

Click here for additional data file. (137.4MB, pptx)

Supporting Information

Click here for additional data file. (422.4KB, docx)

Supporting Information

Click here for additional data file. (20.8KB, docx)

Supporting Information

Click here for additional data file. (15.5KB, docx)

Supporting Information

Click here for additional data file. (16KB, docx)

Supporting Information

Click here for additional data file. (20KB, docx)

Supporting Information

Click here for additional data file. (24.6KB, docx)

Supporting Information

Click here for additional data file. (33.3KB, docx)

Supporting Information

Click here for additional data file. (23.1KB, docx)

Supporting Information

Click here for additional data file. (23.2KB, docx)

Supporting Information

Click here for additional data file. (23.8KB, docx)

Supporting Information

Click here for additional data file. (25.3KB, docx)

Supporting Information

Click here for additional data file. (21.4KB, docx)

ACKNOWLEDGEMENTS

The work was supported by Operation Funding of Shanghai Institute of Clinical Bioinformatics and Shanghai Engineering and Technology Center for Artificial Intelligence of Lung and Heart Diseases from Zhongshan Hospital, National Nature Science Foundation of China (grant number: 81873409), Cross Key Project of Mathematics and Medical Health of National Natural Science Foundation of China (grant number: 12026608) and Pazhou Lab

Contributor Information

Chengshui Chen, Email: wzchencs@163.com.

Yuanlin Song, Email: song.yuanlin@zs-hospital.sh.cn.

Wei Wang, Email: weiwang1@fudan.edu.cn.

Xiangdong Wang, Email: xdwang@fuccb.com.

REFERENCES

  • 1. Jovic D, Liang X, Zeng H, Lin L, Xu F, Luo Y. Single‐cell RNA sequencing technologies and applications: a brief overview. Clin Transl Med. Mar 2022;12(3):e694. 10.1002/ctm2.694 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Peng J, Sun BF, Chen CY, et al. Single‐cell RNA‐seq highlights intra‐tumoral heterogeneity and malignant progression in pancreatic ductal adenocarcinoma. Cell Res. Sep 2019;29(9):725–738. 10.1038/s41422-019-0195-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Papalexi E, Satija R. Single‐cell RNA sequencing to explore immune cell heterogeneity. Nat Rev Immunol. Jan 2018;18(1):35–45. 10.1038/nri.2017.76 [DOI] [PubMed] [Google Scholar]
  • 4. Ortega MA, Poirion O, Zhu X, et al. Using single‐cell multiple omics approaches to resolve tumor heterogeneity. Clin Transl Med. Dec 28 2017;6(1):46. 10.1186/s40169-017-0177-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Butler A, Hoffman P, Smibert P, Papalexi E, Satija R. Integrating single‐cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol. Jun 2018;36(5):411–420. 10.1038/nbt.4096 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Lahnemann D, Koster J, Szczurek E, et al. Eleven grand challenges in single‐cell data science. Genome Biol. Feb 7 2020;21(1):31. 10.1186/s13059-020-1926-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Zheng C, Zheng L, Yoo JK, et al. Landscape of infiltrating T cells in liver cancer revealed by single‐cell sequencing. Cell. Jun 15 2017;169(7):1342–1356 e16. 10.1016/j.cell.2017.05.035 [DOI] [PubMed] [Google Scholar]
  • 8. Dogra P, Rancan C, Ma W, et al. Tissue determinants of human NK cell development, function, and residence. Cell. Feb 20 2020;180(4):749–763 e13. 10.1016/j.cell.2020.01.022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Travaglini KJ, Nabhan AN, Penland L, et al. A molecular cell atlas of the human lung from single‐cell RNA sequencing. Nature. Nov 2020;587(7835):619–625. 10.1038/s41586-020-2922-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Fang H, Zeng Y, Zhang L, Chen C, Powell CA, Wang X. Can single cell RNA sequencing reshape the clinical biochemistry of hematology: new clusters of circulating blood cells. Clin Transl Med. Dec 2021;11(12):e671. 10.1002/ctm2.671 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information

Click here for additional data file. (137.4MB, pptx)

Supporting Information

Click here for additional data file. (422.4KB, docx)

Supporting Information

Click here for additional data file. (20.8KB, docx)

Supporting Information

Click here for additional data file. (15.5KB, docx)

Supporting Information

Click here for additional data file. (16KB, docx)

Supporting Information

Click here for additional data file. (20KB, docx)

Supporting Information

Click here for additional data file. (24.6KB, docx)

Supporting Information

Click here for additional data file. (33.3KB, docx)

Supporting Information

Click here for additional data file. (23.1KB, docx)

Supporting Information

Click here for additional data file. (23.2KB, docx)

Supporting Information

Click here for additional data file. (23.8KB, docx)

Supporting Information

Click here for additional data file. (25.3KB, docx)

Supporting Information

Click here for additional data file. (21.4KB, docx)

Articles from Clinical and Translational Medicine are provided here courtesy of John Wiley & Sons Australia, Ltd on behalf of Shanghai Institute of Clinical Bioinformatics

RESOURCES