OBJECTIVE

To establish the differences and identify the prognostic value of biological markers — platelet growth factors PDGF-AA, PDGF-BB and brain-derived neurotrophic growth factor (BDNF) — for the differential diagnosis of patients with unipolar and bipolar depression using machine learning.

MATERIAL AND METHODS

The study included 79 patients aged 48 [34; 57] years, including 35 people with bipolar depression (ICD-10. F31) and 44 people with unipolar depression (F32-33). Clinical assessment of the patients’ condition was carried out using the Hamilton Depression Rating Scale (HDRS-17) and the Hamilton Anxiety Rating Scale (HARS). The concentration of growth factors in the blood serum of patients was determined using multiplex analyzers Magpix and Luminex 200 (Luminex, USA). To build a predictive model, the support vector machine was used.

RESULTS

Patients with bipolar depression showed statistically significant higher concentrations of PDGF-AA and PDGF-BB, as well as lower concentrations of BDNF. When constructing a predictive model, it was possible to separate patients with unipolar and bipolar depression according to all three biomarkers; the sensitivity and specificity of the model were 0.96±0.06 and 0.95±0.05, respectively.

CONCLUSIONS

The study of concentrations of BDNF and platelet-derived growth factors shows statistically significant differences in indicators in the case of unipolar and bipolar depression, which can potentially be used as prognostic biomarkers for differential diagnosis in appropriate clinical cases.