Ai-assisted Integrative Transcriptomic Analysis Supports Mechanistic Delineation Of Nodular And Extensive Keloid Subtypes With Therapeutic Implications
Kevin Serror 1,2,3*°, Lina Faik 4°, Thaïs Hennebelle 1,2, Xavier Sécheresse 4, Léa Nicosia 1,2, Pierre De La Grange 5, David Boccara 3, Marc Chaouat 3, Jean-David Bouaziz 1,2,6, Joël Belafa 4, Maxime Gobin 4, Antoine De Torcy 4, Florence Armstrong 4#, and Laurence Michel 1,2,6*
BACKGROUND
Keloids are heterogeneous fibrotic skin disorders in which nodular and extensive forms differ in growth behavior, recurrence, and treatment response. While transcriptomic studies have identified subtype-specific signatures, the mechanistic architectures underlying these divergent phenotypes remain unresolved.
OBJECTIVE
To achieve a system-level dissection of nodular and extensive keloids to support subtype-adapted treatment using AI-assisted transcriptomic analysis.
METHODS
We reanalyzed published transcriptomic data from primary dermal fibroblasts of nodular, extensive keloids and healthy skin using OCEAN (Omics-to-Clinical Evidence & Action Navigator), an AI-assisted framework integrating differential expression, pathway enrichment, and protein–protein interaction networks into coherent mechanistic architectures with traceable evidence. Cultured fibroblasts from normal skin biopsies or extensive/nodular keloids were used for biological experiments aimed to validate IA underlined pathways and new pharmacological target molecules.
RESULTS
AI-assisted analysis confirmed known subtype differences, including a dominant extracellular matrix program in nodular keloids and a more inflammatory profile
in extensive keloids. System-level integration reconstructed distinct mechanistic
architectures: nodular keloids displayed a compact, self-reinforcing fibrotic network integrating TGF-β signaling, hypoxia adaptation, transcriptional stabilization, and autocrine inflammation, consistent with reduced therapeutic responsiveness. In contrast, extensive keloids showed an adaptive, expansion-oriented program marked by immune responsiveness, stromal plasticity, angiogenic signaling, and metabolic stress adaptation, revealing subtype-specific therapeutic vulnerabilities.
CONCLUSION
AI-assisted integrative transcriptomics transforms descriptive omics profiles into clinically actionable mechanistic insight, enabling precision stratification and mechanism-informed therapeutic targeting in keloid disease.
METRICS
Kevin Serror 1,2,3*°, Lina Faik 4°, Thaïs Hennebelle 1,2, Xavier Sécheresse 4, Léa Nicosia 1,2, Pierre De La Grange 5, David Boccara 3, Marc Chaouat 3, Jean-David Bouaziz 1,2,6, Joël Belafa 4, Maxime Gobin 4, Antoine De Torcy 4, Florence Armstrong 4#, and Laurence Michel 1,2,6*#
- Inserm UMRS_1342, Hôpital Saint-Louis, Paris, France
- Université Paris Cité, Paris, France
- Service de Chirurgie plastique, reconstructive et esthétique, Hôpital Saint-Louis, Paris, France
- Biolevate, Paris, France
- GenoSplice, Paris, France
- Service de Dermatologie, APHP, Hôpital Saint Louis, Paris, France °equal contribution # Co-seniors
Running Title
AI-assisted integrative transcriptomic analysis supports mechanistic delineation of nodular and extensive keloid subtypes with therapeutic implications
Word Count
226 words