Related Experiment Video
Updated: Jan 7, 2026
Isolation, Characterization and MicroRNA-based Genetic Modification of Human Dental Follicle Stem Cells
Published on: November 16, 2018
Predictive modeling of stem cell suppression by inflammatory cytokine networks: A synthetic transcriptomic approach
Deepavalli Arumuganainar1, Pradeep Kumar Yadalam2
1Department of Periodontology, Saveetha Dental College and Hospitals, Saveetha Institute of Medical and Technical Sciences (SIMATS), Chennai, Tamil Nadu, India.
Background:
Chronic inflammation in periodontitis disrupts the osteogenic differentiation of periodontal ligament stem cells (PDLSCs) through persistent cytokine activity. IL-1β, TNF-α, and IL-6 are major mediators that inhibit bone-forming pathways. However, the complexity of cytokine-gene interactions remains poorly characterized. This study presents a synthetic transcriptomic modeling framework to predict and interpret inflammatory suppression of stem-cell osteogenesis.
Methods:
Time-resolved synthetic gene expression profiles were generated to simulate osteogenic induction under homeostatic, inflammatory, and resolution phases. A curated gene regulatory network (GRN) was incorporated to map cytokine-osteogenesis interactions. Graph autoencoders (GAEs) captured latent topological structure from the expression matrix, while deep neural classifier differentiated inflammatory from control states. The GSE283726 periodontitis transcriptome dataset and iPSC-derived mesenchymal stem cells (iMSCs) were used for validation.
Results:
Simulations showed that IL-1β and TNF-α strongly activated NF-κB signaling, suppressing osteogenic genes such as RUNX2 and ALPL. IL-6 exhibited context-dependent regulatory behavior. GAEs clearly separated inflammatory and regenerative modules, identifying IL-6 as a key intermediary. The classifier achieved an AUROC of 0.99 and > 95 % accuracy. Validation with real datasets confirmed overlap in differentially expressed genes and enriched pathways, including Wnt inhibition (DKK1) and inflammatory GO terms.
Conclusion:
Biologically informed synthetic transcriptomics combined with graph autoencoding effectively models cytokine-mediated inhibition of PDLSCs. The framework identifies regulatory nodes supported by real data and offers potential for in silico drug testing. Future work will expand cytokine networks, incorporate diverse cell types, and explore transfer learning for regenerative periodontal applications.
