Tang et al. 2026: Identification of Signature Genes in Type 2 Diabetes

Study Design

• scRNA-seq source: GSE221156 — pancreatic islet cells from 7 T2D, 17 non-diabetic, and 14 prediabetic individuals.
• Independent validation: GSE29221 — skeletal muscle microarray from 3 T2D and 3 non-diabetic males.
• Clinical validation: qRT-PCR on peripheral blood serum from 15 T2DM patients and 20 healthy controls (Shenzhen Third People’s Hospital, single-center, Chinese cohort).
• Primary goal: Systematic screening of T2D signature genes by integrating scRNA-seq with machine learning.

Key Methods

• scRNA-seq pipeline: Seurat 4.0, Harmony batch correction, SingleR annotation → 10 cell types from 19 subclusters.
• DEG identification: 455 DEGs via limma.
• LASSO regression: 14 genes selected from 455 DEGs (lambda.1se = 0.093, 5-fold CV, seed 666).
• Overlap: 4 genes common to both DEGs and LASSO selection: PNLIP, BUB1, CTSB, NAMPT.
• Functional enrichment: GO/KEGG via clusterProfiler.
• Cell-cell communication: CellChat (MK and SPP1 pathways highlighted).

Key Findings

Signature Genes and Diagnostic Performance

Gene	AUC (GSE29221)	qRT-PCR Direction	Known Function
BUB1	0.931	Up in T2D	Mitotic checkpoint kinase; cell cycle regulation
CTSB	0.882	Up in T2D	Cathepsin B; lysosomal cysteine protease
PNLIP	0.819	Up in T2D	Pancreatic lipase; triglyceride digestion
NAMPT	0.694	Down in T2D	NAD+ biosynthesis rate-limiting enzyme; also pro-inflammatory eNAMPT cytokine

Functional Enrichment

Functional enrichment of the 455 DEGs (clusterProfiler) revealed processes spanning extracellular remodeling, digestive function, and signaling:

GO Biological Process:

• Collagen metabolism and extracellular matrix organization — consistent with pancreatic islet fibrosis, a known T2D pathology.
• Serine-type endopeptidase activity — related to protease signaling and zymogen activation in pancreatic acinar cells.

GO Cellular Component:

• Endoplasmic reticulum lumen — reflects ER stress in T2D islet cells, particularly beta cells under secretory demand.
• Extracellular region and extracellular exosome — secreted factors and matrix components.

KEGG Pathways:

• Fat digestion and absorption — pancreatic lipase related (PNLIP), links digestive function to metabolic dysregulation.
• Pancreatic secretion — aligns with exocrine pancreatic involvement in T2D.
• Protein digestion and absorption — cathepsin related (CTSB), consistent with extracellular proteolysis.
• Neuroactive ligand-receptor interaction — suggests islet neuronal or receptor-mediated signaling changes.

Among the four signature genes, PNLIP drove the digestive/metabolic enrichment, CTSB contributed to proteolysis and ECM categories, and NAMPT linked to NAD metabolism and inflammatory signaling. BUB1’s mitotic function was less represented in the DEG-level enrichment but reflects compensatory proliferation or cell-cycle stress in T2D islets.

Cell Communication (CellChat)

• Alpha and Beta cells identified as signaling hubs within the islet microenvironment.
• MK (midkine) and SPP1 (osteopontin) pathways showed complementary expression patterns across cell types.
• CTSB was linked to SPP1-mediated signaling, consistent with its role in extracellular proteolysis and tissue remodeling.
• The authors note that NAMPT forms receptor-ligand pairs implicated in diabetic kidney disease (citing external literature), but this was not tested in their dataset.

Clinical Validation

• qRT-PCR confirmed PNLIP, BUB1, CTSB significantly upregulated; NAMPT significantly downregulated in T2DM serum (p < 0.01).

Methodological Notes

• scRNA-seq source is pancreatic islet, not PBMC — this study captures tissue-level pathology rather than systemic immune remodeling.
• ROC validation dataset (GSE29221) is small (n=6, skeletal muscle) — a cross-tissue validation that limits generalizability.
• Clinical validation is single-center, modest sample size (n=35 total), and from a Chinese population — ancestry-specific effects not testable.
• Authors acknowledge these findings are hypothesis-generating and require larger cohorts and functional studies.

Relevance to This Paper

• Provides T2D biomarker candidates from islet tissue with blood-derived qRT-PCR validation — a cross-compartment translation approach.
• Complements Huang et al. 2022 as another islet-derived biomarker study with a different gene panel and methodology.
• Does not test ancestry-associated differences; all participants were from a single Chinese hospital.
• The scRNA-seq + ML pipeline approach is methodologically similar to several recent T2D biomarker studies.

Links

• GSE221156 — pancreatic islet scRNA-seq
• GSE29221 — skeletal muscle microarray validation
• Type 2 Diabetes
• T2D Islet Biomarkers
• T2D Islet Transcriptome Biomarker Discovery
• Huang et al. 2022 × Tang et al. 2026 — synthesis