Vildagliptin for skin fibrosis

Can a diabetes drug quiet the cells that harden skin? A seven-day repurposing study, tested layer by layer.

Role: Co-led analysis, pitch
Date: 2026-08
Status: live
evidence: 7 layers
cohorts: 4 datasets

DPP4-positive cells, %
Cell	control	SSc
Fibroblast	9.58	14.65
NK	2.84	3.43
Endothelial	1.72	1.22
Myeloid	1.21	0.85
Mast	0.55	1.14
Keratinocyte	0.39	0.66
T cell	0	0.47
Pericyte	0.03	0

Fig. 1. Fibroblast states in systemic sclerosis skin (GSE138669), connected by PAGA. Colour is the SSc effect on each state’s share of cells. The DPP4-high branch runs 0 → 2 → 4. Values are read off the published figure, so they’re approximate. PAGA shows how states connect; on its own it says nothing about which way a lineage runs.

Vildagliptin blocks the enzyme DPP-4. In systemic sclerosis, DPP-4 marks the progenitor fibroblasts that become collagen-producing myofibroblasts. At the Cortellis Drug Repositioning Hackathon, a team of five tested whether that overlap justifies a lab experiment.

The evidence came from four public human skin cohorts, single-cell and network analyses, and a pharmacokinetic model of how much drug reaches skin. Some layers supported the idea and some narrowed it: DPP4 is not globally increased and is not a severity marker. The verdict was to advance to a proof-of-mechanism experiment, which was designed but has not been run.

I co-led the computational analysis and presented the case. The project page re-plots every result as an interactive chart, and each of its 54 references is checked against PubMed.