Objective — to use Bayesian network (BN) technology to predict disease progression and death in patients with breast cancer (BC) in relation to both traditional risk factors used in oncology and the expression of the new molecular prognostic factor — multifunctional protein YB-1. Material and methods. The database including patients with a verified diagnosis of Stages I—IV BC was built on the basis of information on 32 clinical and molecular biological parameters in 323 patents. The endpoints (EP) were taken to be disease progression (EP1-P) and death (EP2-D) in the first four years after surgical treatment. Its BC with naïve topology was constructed for each EP. Results. To improve prediction quality, the investigators optimized the baseline BNs by the number of nodes (the latter were removed one by one until the value of AUC, the area under the ROC curve built on the database by excluding patients one by one, increased). The same fashion was used to find the optimal BNs containing 7 nodes each and having high-quality prediction: AUC=0.8331 for EP1P and 0.9073 for EP2-D. The expression of YB-1 mRNA in tumor tissue was among 7 prognostic parameters determining the risk of death. The optimal BNs were employed to construct risk histograms to stratify patients by risk groups and to compute the probability of respective outcomes. Conclusion. The use of a BN with naïve topology was shown to be effective in revealing the parameters critical for the prediction of the course of BC in patients; the optimal BNs with high-quality prediction were found; risk histograms that might be used in the expert system of personified medicine were also created.