1. INTRODUCTION
The human pancreatic islet of Langerhans consists of five main endocrine cell types, collectively regulating nutrient homeostasis and metabolism. Single‐cell RNA sequencing (scRNAseq) of the pancreas revealed how factors such as age, sex, body mass index (BMI), and disease state affect islet cell transcriptomes. 1 , 2 , 3 The Human Pancreas Analysis Program (HPAP) has generated rich human data from non‐diabetic and diabetic donors, including pancreatic scRNAseq, demographic, clinical, histological, and physiological data. 4 , 5 , 6 Data are freely available, but its magnitude and complexity make it difficult to analyse without training in bioinformatics.
2. METHODS
We designed a web‐based research tool for non‐bioinformatician diabetes and endocrine researchers and physicians that enables gene expression analysis across subsets of patients and endocrine cells curated by HPAP. The tool provides information on the number of cells analysed and metadata on the patient donors included in the analysis (https://singlecellrnapancreas.shinyapps.io/DBZvi/). Our goal was to keep the tool simple, to facilitate quick and user‐friendly analysis for the most common queries.
We optimized the analysis for endocrine cells, which are the research focus of most scientists and clinicians in the field, and since the human islet isolation protocol is inherently suboptimal for isolation of immune and exocrine/duct cells, reducing the quality of the data and generating potential biases. We performed rigorous data quality control with parameters tailored to the unique expression pattern of pancreatic endocrine cells, which predominantly express the main hormone secreted (e.g., INS in β‐cells), and the statistics of the HPAP scRNAseq protocols. 7 Our stringent approach retained only cells of high transcriptomic quality. The data include 80 donors and 139 946 endocrine cells. The extensive HPAP data set offers the opportunity to test gene expression hypotheses in populations that can be retrospectively controlled for demographic, medical, and technical parameters. A user guide, technical details, and code are provided on the website.
The research tool has two main parts. First, the researcher can select a subset of the patients according to demographic parameters (sex, age, ethnicity), clinical parameters (BMI, HbA1C, disease type, autoantibody detection, disease duration if known), or choose to include/exclude specific donors. The researcher can also select specific cell types according to our annotations or filter according to the mRNA library protocol used in HPAP. This selection enables for example matching by age when comparing α‐cells from non‐diabetic and type 2 diabetes mellitus (T2DM) patients, as T2DM donors are usually older. The metadata and number of cells of the selected donors can be viewed using summary tables and histograms, to detect biases in the data in sex, ethnicity, BMI, and number of cells. The researcher can select to analyse cells that express a gene at a level greater than some threshold.
After the selection of the donors and cells, the researcher can perform gene expression analyses. The expression of a gene of interest can be viewed in each cell type using one or two categories (e.g., age and/or HbA1C). The violin or box plots show both the percentage of cells that express the gene of interest and the distribution of gene expression. Since many genes are detected in a low percentage of cells, we also generate a violin/box plot of gene expression only in cells in which the transcript is detected. We test the null hypothesis that the meta‐data parameter chosen does not affect the distribution of gene expression, based on the rank‐sum statistics. Importantly, the resulting p‐value is not corrected for multiple hypothesis testing or confounding factors, and our test considers all cells as independent samples. These assumptions are not necessarily valid in all cases. The statistical output, descriptive statistics such as mean and median expression levels, and the number of cells and patients in each group can be viewed and downloaded.
More experienced researchers can visualize the expression of a gene of interest in a uniform manifold approximation and projection (UMAP) plot, and colour‐code the data according to each of the meta‐data parameters and further sub‐select donors according to a range of age, BMI, or HbA1C to determine whether the selected subset is mapped to a different region in the UMAP. Finally, the user can also perform a differential expression analysis between two subgroups of cells.
Below are two examples of quick analyses of gene expression along with screenshots from the tool.
3. RESULTS
3.1. TXNIP expression in T2DM and non‐diabetic donors
Elevated expression of the metabolic regulator TXNIP is associated with the pathogenesis of T2DM. 8 Here, we study TXNIP expression in β‐cells in non‐diabetic and T2DM donors. After selecting the donors and cells, we remain with 49 039 cells from 68 donors: 74% are non‐diabetic, 55% are females, half have obesity (BMI >30 kg/m2), and 56% are Caucasian. Studying a plot of the donor metadata is recommended prior to any analysis to detect biases in the data (Figure 1A).
FIGURE 1.
Screenshots from the web tools showing TXNIP expression in cells of type 2 diabetes mellitus (T2DM) and non‐diabetic donors. Fonts were enlarged for visualization in a figure panel. (A) Distribution of donors by frequency of HbA1C ranges, BMI ranges, age ranges, sex, disease condition, and ethnicity. (B) Box‐plot of TXNIP expression in β‐cells of non‐diabetic donors (left, pink) and donors who had T2DM (right blue). The percentage of TXNIP‐positive cells and outcome of a non‐parametric rank‐sum test is provided above the plot. (C) Box plot as in (C), across age ranges. Note that there were no donors with T2DM younger than 30 years old. (D, E) Table of the number of donors (D) and number of cells (E) across age groups and conditions in the analysis in (A–C). (F) Box plot as in C comparing TXNIP expression in donors aged 41–60 who had T2DM or that were non‐diabetic in alpha, beta, delta, gamma, and epsilon cells.
As expected, TXNIP is detected in a higher level and a higher fraction of β‐cells of T2DM donors compared with non‐diabetic donors (Figure 1B). Since donors with T2DM are usually older than non‐diabetic donors, we compared TXNIP expression across age and disease condition of the donors (Figure 1C). There are no diabetic donors younger than 31 years, creating a bias in the previous analysis; yet the data show higher TXNIP expression in donors with T2DM in each age group. The tool provides the number of donors and cells analysed in each age group and condition revealing that most T2DM data were provided from donors aged 41–60, calling for greater caution in interpreting the data in a wider age range (Figure 1D,E). We tested if TXNIP expression in other cell types was affected by the diagnosis of T2DM, and limited our analysis to ages 41–60. To do so, we reset our cell selection and included this specific donor population and all cell types, resulting in 59 198 cells. We observe an increase in TXNIP expression in this age group in α‐, β‐, δ‐ and γ‐cells in T2DM donors compared to non‐diabetic donors (Figure 1F) providing preliminary evidence that TXNIP upregulation in T2DM is not unique to β‐cells.
3.2. HLA‐E expression in α‐ and β‐cells of type 1 diabetes mellitus and non‐diabetic donors
HLA‐E is an immunoregulatory protein associated with α‐cell protection against the autoimmune assault in type 1 diabetes mellitus (T1DM). 9 We selected α‐ and β‐cells of the 46 non‐diabetic and 12 T1DM donors (93 631 cells). A UMAP plot shows high variability in HLA‐E expression, with higher expression in α‐cells (Figure 2A,B). T1DM donors have higher HLA‐E expression than non‐diabetic donors in β‐ and α‐cells (Figure 2C). Different ethnicities had marked differences in HLA‐E expression in both T1DM and non‐diabetic donors (Figure 2D). Notably, while there are at least five donors in each ethnicity group among the non‐diabetic donors, there is only a single Hispanic and single African American donor that were diagnosed with T1DM, limiting interpretability of the results in the T1DM population (Figure 2E). Non‐diabetic female donors display mildly lower HLA‐E expression, while T1DM female donors have higher HLA‐E expression than males (Figure 2F), with a balanced distribution across sex (Figure 2G). Exploratory differential expression analysis comparing gene expression in α‐cells of male and female T1DM patients showed surprisingly that CHGB is one of the most highly upregulated genes in females compared to males. We visualize CHGB expression in T1DM donors using a violin plot (Figure 2H), validation of this finding at the protein level is warranted.
FIGURE 2.
Screenshots showing HLA‐E expression in α‐ and β‐cells of type 1 diabetes mellitus (T1DM) and non‐diabetic donors. Fonts were enlarged for visualization in a figure panel. (A) Uniform manifold approximation and projection (UMAP) of HLA‐E expression in α‐ and β‐cells. (B) Violin plot of HLA‐E expression level in α‐cells (left) and β‐cells (right) of non‐diabetic (pink) and T1DM (blue) donors. (C) Violin plot of HLA‐E expression level in α‐cells (left) and β‐cells (right) categorized by female (pink) and male (blue) non‐diabetic and T1DM donors. (D) HLA‐E expression level in α‐cells (left) and β‐cells (right) categorized by ethnicity: African American (pink), Caucasians (green), and Hispanic (blue) non‐diabetic and T1DM donors. (E) Number of non‐diabetic and T1DM donors by ethnicity. (F) HLA‐E expression level in α‐cells (left) and β‐cells (right) categorized by sex: females (pink), and males (blue) in non‐diabetic and T1DM donors. (G) Number of non‐diabetic and T1DM donors by sex. (H) CHGB expression in alpha cells of female (pink) and male (blue) T1DM donors.
4. CONCLUSIONS
We present a user‐friendly, simple yet versatile tool for the analysis of scRNAseq data of the human endocrine pancreas generated by the HPAP consortium. Initiatives commonly used in cell atlases to analyse scRNAseq data. Our platform is different from the existing platforms used to analyse scRNAseq HPAP data such as PANC‐DB, CellxGene and Azimuth 10 , 11 , 12 in its optimization for endocrine cell types, in its simple ability for nested analyses, presentation of biases due to donor metadata, and in aiming for researchers that find it challenging to interpret data using UMAPs and advanced methods. This tool does not support integration of other HPAP data such as histology and proteomics and is limited in its and statistical modelling, and does not replace formal statistical analyses and validation by other data sets and experimental methods.
Our platform invites the researcher to perform simple analyses of gene expression in multiple, nested subsets of donors and cells and view the data in graphic or tabular form with simple statistical analyses. We hope it will enable non‐bioinformatician researchers to easily access and explore high‐quality transcriptomics data to quickly test hypotheses.
AUTHOR CONTRIBUTIONS
HB, RBHS and DBZ designed the website. CB, XY, RBHS, DLE and DBZ performed analyses and wrote the manuscript.
FUNDING INFORMATION
DBZ: Zuckerman STEM faculty, ISF‐JDRF 1541/21. DBZ is the Gad Avigad Chair in Biochemistry. DLE is supported by grants from Breakthrough T1D (3‐SRA‐2022‐1201‐S‐B), 1 , 2 the NIH Human Islet Research Network Consortium on Beta Cell Death and Survival (HIRN‐CBDS, U01 DK127786) and NIH/NIDDK grants RO1DK126444 and RO1DK133881‐01.
CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.
Benhayon H, Yi X, Schyr RB‐H, Eizirik DL, Ben‐Zvi D. A web tool for easy and versatile analysis of human endocrine pancreas single‐cell RNAseq data. Diabetes Obes Metab. 2026;28(2):1545‐1549. doi: 10.1111/dom.70281
DATA AVAILABILITY STATEMENT
This manuscript used data acquired from the Human Pancreas Analysis Program (HPAP‐RRID:SCR_016202) Database (https://hpap.pmacs.upenn.edu/) and Human Islet Research Network (RRID:SCR_014393) consortium (UC4‐DK‐112217, U01‐DK‐123594, UC4‐DK‐112232 and U01‐DK‐123716). The code is available through the readme section on the website (https://singlecellrnapancreas.shinyapps.io/DBZvi/).
REFERENCES
- 1. Yong HJ, Toledo MP, Nowakowski RS, Wang YJ. Sex differences in the molecular programs of pancreatic cells contribute to the differential risks of type 2 diabetes. Endocrinology. 2022;163(11):bqac156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Shrestha S, Erikson G, Lyon J, et al. Aging compromises human islet beta cell function and identity by decreasing transcription factor activity and inducing ER stress. Sci Adv. 2022;8:3932. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Ewald JD, Lu Y, Ellis CE, et al. HumanIslets.com: improving accessibility, integration, and usability of human research islet data. Cell Metab. 2025;37:7‐11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Shapira SN, Naji A, Atkinson MA, Powers AC, Kaestner KH. Understanding islet dysfunction in type 2 diabetes through multidimensional pancreatic phenotyping: the Human Pancreas Analysis Program. Cell Metab. 2022;34:1906‐1913. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Kaestner KH, Powers AC, Naji A, Atkinson MA. NIH initiative to improve understanding of the pancreas, islet, and autoimmunity in type 1 diabetes: the Human Pancreas Analysis Program (HPAP). Diabetes. 2019;68:1394‐1402. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Patil AR, Schug J, Naji A, Kaestner KH, Faryabi RB, Vahedi G. Single‐cell expression profiling of islets generated by the Human Pancreas Analysis Program. Nat Metab. 2023;5:713‐715. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Hao Y, Stuart T, Kowalski MH, et al. Dictionary learning for integrative, multimodal and scalable single‐cell analysis. Nat Biotechnol. 2023;42(2):293‐304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Minn AH, Hafele C, Shalev A. Thioredoxin‐interacting protein is stimulated by glucose through a carbohydrate response element and induces beta‐cell apoptosis. Endocrinology. 2005;146:2397‐2405. [DOI] [PubMed] [Google Scholar]
- 9. Colli ML, Ramos‐Rodríguez M, Nakayasu ES, et al. An integrated multi‐omics approach identifies the landscape of interferon‐α‐mediated responses of human pancreatic beta cells. Nat Commun. 2020;11:11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Elgamal RM, Kudtarkar P, Melton RL, et al. An integrated map of cell type–specific gene expression in pancreatic islets. Diabetes. 2023;72:1719‐1728. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. CZI Cell Science Program , Abdulla S, Aevermann B, et al. CZ CELLxGENE discover: a single‐cell data platform for scalable exploration, analysis and modeling of aggregated data. Nucleic Acids Res. 2025;53:13‐14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Hao Y, Hao S, Andersen‐Nissen E, et al. Integrated analysis of multimodal single‐cell data. Cell. 2021;184:3573‐3587.e29. [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.
Data Availability Statement
This manuscript used data acquired from the Human Pancreas Analysis Program (HPAP‐RRID:SCR_016202) Database (https://hpap.pmacs.upenn.edu/) and Human Islet Research Network (RRID:SCR_014393) consortium (UC4‐DK‐112217, U01‐DK‐123594, UC4‐DK‐112232 and U01‐DK‐123716). The code is available through the readme section on the website (https://singlecellrnapancreas.shinyapps.io/DBZvi/).