Machine Learning-Based Transcriptomic Diagnosis of Periodontitis.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Translational Medicine

Background:

• Periodontitis diagnosis relies on subjective and low-sensitivity conventional methods, posing a global health challenge.
• Transcriptomic data offers a potential avenue for developing more accurate and efficient diagnostic tools.

Purpose of the Study:

• To develop and validate a machine learning (ML)-based diagnostic framework for periodontitis using transcriptomic data.
• To identify novel gene biomarkers and understand their role in periodontitis pathogenesis.

Main Methods:

• Analysis of transcriptomic datasets from 616 samples (periodontitis and healthy controls).
• Identification of differentially expressed genes (DEGs), functional enrichment, WGCNA, and immune infiltration profiling.
• Biomarker refinement using Boruta and LASSO, followed by ML model construction and validation (RF, XGBoost).

Main Results:

• Five diagnostic biomarkers (CSF2RB, COL15A1, MME, NEFM, CYP24A1) were identified with high accuracy in training and validation datasets.
• ML models, particularly Random Forest and XGBoost, demonstrated excellent classification performance.
• Significant correlations between biomarkers and immune cell populations were observed, alongside identification of key regulatory transcription factors (NFYA, SP1).

Conclusions:

• An ML-driven diagnostic framework integrating transcriptomic, immune, and regulatory network insights was established for periodontitis.
• The identified gene biomarkers and diagnostic models show potential for clinical utility in personalized diagnosis, targeted intervention, and therapeutic development.