T2D Research

Huang et al. 2022 × Tang et al. 2026

Properties2
tagssynthesis, type-2-diabetes, transcriptomics, biomarker, pancreatic-islet
aliasesHuang et al. 2022 × Tang et al. 2026

3 min read

Huang et al. 2022 × Tang et al. 2026

The Connection

Both studies use pancreatic islet transcriptomics and LASSO regression to identify candidate diagnostic T2D gene panels, but they start from different data types — Huang et al. uses bulk RNA-seq DEGs as the LASSO input, while Tang et al. uses scRNA-seq DEGs. The resulting gene panels share no members, despite both targeting T2D and pancreatic-islet biology. This methodological divergence despite related biological context is the central synthesis insight: islet-derived T2D biomarker candidates may be method- and dataset-dependent.

Where They Co-occur

Both references co-occur in the T2D islet biomarkers concept page, the islet transcriptome biomarker discovery concept page, the type-2-diabetes hub, the project page, and the evidence map. Their 5 co-occurring pages reflect the paper’s interest in islet biology as complementary tissue context for T2D pathology.

Cross-cutting Insight

The non-overlapping gene panels tell a cautionary tale for biomarker discovery: Huang et al.’s bulk-first approach identifies glucose-transport, serpin, and adhesion-molecule candidates with islet-context plausibility, while Tang et al.’s scRNA-seq-first approach identifies cell-cycle, protease, lipase, and NAD-biosynthesis candidates with limited cross-compartment expression checks. Neither panel is necessarily wrong — they may capture different biological dimensions of islet pathology because starting data type, cohort, cell-type resolution, and validation design shape which genes are detectable as differentially expressed. For the ancestry paper, this suggests that PBMC-derived biomarker candidates may be equally method-dependent, and cross-platform validation is essential before any panel is treated as robust.

Tensions and Trade-offs

  • Huang et al. report within-dataset nomogram performance (AUC 0.902) without independent cross-tissue testing, while Tang et al. check expression across islet, muscle, and peripheral blood/serum. These are different validation strategies, not equivalent clinical validation.
  • Tang et al.’s qRT-PCR in blood tests whether islet-derived candidates are detectable/differential in a small single-center Chinese peripheral blood/serum cohort (n=35), limiting generalizability and not establishing PBMC immune relevance.
  • Huang et al. also mapped where their biomarkers localize within islet cell types and which transcription factors regulate them. Tang et al. did not do this, but did validate their panel across three different tissues. Neither approach is complete on its own — they work best together.
  • For manuscript use, both studies support T2D biomarker plausibility from islet biology but neither tests PBMC or ancestry-specific performance — they are background context, not direct evidence for the paper’s central claim.

Open Questions

  • Would either panel perform as PBMC biomarkers if tested in a multi-ancestry cohort with T2D phenotyping?
  • Do any of the eight genes (SLC2A2, SERPINF1, RASGRP1, CHL1, PNLIP, BUB1, CTSB, NAMPT) show ancestry-associated expression variation in PBMC data?
  • How sensitive are LASSO-selected T2D biomarker panels to the choice of starting DEG list (bulk vs. single-cell, different thresholds)?

Graph View

Backlinks