НЕЙРОСЕТЕВОЙ АНАЛИЗ НЕКОТОРЫХ МОЛЕКУЛЯРНЫХ ПАРАМЕТРОВ ЦЕРВИКАЛЬНОГО ЭПИТЕЛИЯ ДЛЯ ДИАГНОСТИКИ РАКА ШЕЙКИ МАТКИ

Background. A personalized approach is the basis for the specialized care for cancer patients. The relevance of cervical cancer (CC) is still high. The searches for diagnostic criteria of cervical epithelium malignancy are continuing. The application ohm technologies has led to a big number results, the analysis of which is often difficult. The neural network data analysis allows to solve these problems. Objective : to create a technology for diagnosing cervical intraepithelial neoplasia (CIN) and CC, based on a neural network analysis of some molecular parameters. Materials and methods . The research carried out among patients with CIN III (n = 15), patients with CC stages I–IV (n = 49). The control group consisted of female volunteers without cervical pathology (n = 15). Studied molecular parameters: the spectrum of fatty acids was determined in cervical biopsies, proteins OPN, ICAM-1 were studied in blood serum, proteins of the immune cycle sCD25, sCD27 – in the cervical epithelium. Research methods: gas-liquid chromatography, flow cytometry. Results. Significant differences of fatty acids spectrum, local level sCD27 were revealed in among the studied groups. The multilayer perceptron included C18:2ω6, OPN, ICAM-1, sCD25, sCD27. The performed neural network analysis of the molecular data allows to diagnose CIN III (Se = 0.92; Sp = 0.87; AUC = 0.94; p˂0.001) and CC (Se = 1.00; Sp = 1.00; AUC = 1.00; p˂0.001). Conclusion. The created model makes it possible to diagnose CIN III and CC with high accuracy. The configuration of the multilayer perceptron allows confirming the pathophysiological relationships between the studied molecular parameters, to expand the understanding of the mechanisms of cervical carcinogenesis.