Artificial intelligence in the immunodiagnostics of chronic periodontitis

Artificial intelligence is used to diagnose various diseases of the oral cavity. In the field of clinical laboratory diagnostics, machine learning algorithms are used in the interpretation of complex biochemical data. The purpose of this study was to search for significant infectious-immunological clinical and laboratory data based on a machine learning algorithm for chronic periodontitis. To do this, 124 patients aged 40 to 70 years diagnosed with chronic periodontitis were examined by real-time PCR to detect the periodontal pocket DNA of human herpes viruses and bacterial periodontopathogenic microflora Fusobacterium nucleatum, Treponema denticola, Porphyromonas endodontalis etc., and Porphyromonas gingivalis. Matrix RNAs of proinflammatory cytokines and other markers of chronic inflammatory process were also studied: IL-1β, IL-10, IL-18, TNFa, TLR4, GATA3, CD68. TNFa, IFNg, IL-1β, IL-4, IL-6, IL-10, IL-18; VEGF were determined in a dentoalveolar fluid. Immune cells of the oral cavity were evaluated by analyzing level of CD3⁺, CD4⁺, CD8⁺, CD3⁺HLA-DR⁺, CD64⁺16⁺14^–, CD4⁺25⁺127^+low, CD3⁺CD16⁺CD56⁺, CD3^–CD16⁺CD56⁺, CD14⁺, CD14⁺HLA-DR⁺, CD19⁺HLA-DR⁺, CD19⁺CD5⁺B27^–, CD19⁺CD5^–B27^–, CD19⁺CD5^–B27⁺ cells. Random forest machine learning was used to evaluate the data. A relationship between pathogenic microflora and modality of immune response was revealed. The proinflammatory component reflected in the expression of IL-1β, TNFa, and IFNg mRNA, prevailed in the immune response against aggressive periodontal pathogens: T. denticola, F. nucleatum, etc. The random forest machine learning algorithm selected correlation ratios r ≥ 0.5 (both positive and negative) from a set of data for further analysis by the operator. The random forest machine learning model showed the following significant combinations of data by 10% with a teacher: VEGF, CD3⁺, CD14⁺HLA-DR, CD19⁺CD5^–CD27⁺, as well as TLR4, IL-1b, IL-10, TNFa, and IL-18 mRNA. The development of the applied “random forest” machine learning model with a teacher has already shown a 25% difference: P. endodontalis, GATA3, CD3⁺, CD14⁺, CD19⁺CD5^–CD27⁺, as well as TLR4, TNFa, IL-1b, IL-10, and IL-18 mRNA. The search for significant infectious-immunological clinical and laboratory data based on a machine learning algorithm for chronic periodontitis has shown the importance of proinflammatory cytokines, monocytes, T-lymphocytes and memory B-cells in the development of osteodestructive inflammatory process of mRNA to reveal non-evident causality factors.