Background: The goal of this studywas to select the key factors affecting choice between radical nephrectomy (RN) and partial nephrectomy (PN) for patients with localized RCC based on clinical and nephrometry data. Methods: A special retrospective cohort study was conducted in National Cancer Institute of Ukraine, which results were further validated on patient dataset in urological department of University Clinic of Cologne. The Institutional Review Boards and the local ethics committees of both high-volume centres approved the study. The main nephrometry parameters of tumor location in the kidney were analysed according to the R.E.N.A.L nephrometry score. The remaining functional parenchymal volume (RFPV) was calculated using the special formula. To determine the relationship between the risk of RN or PN, the multivariate predictive modelling method containing 12 parameters was used (Artificial Neural Networks [ANN]). Data validation based on referential centre experience using ROC-curve analysis to detect clinical applicability of the null hypothesis was performed. Results: Based on the analysis, for polary and laterally located tumors, the risk of RN was conditioned only by RFPV. The average critical value of RFPV for polar lesions was X6_crit = 58% (in X6 < X6_crit, RN was predicted); for lateral tumors - X6_crit = 67% (in X6 < X6_crit, RN was predicted). For medial location, the risk of RN only depended on the tumor size. Average critical value of the tumor size in the medial location was X7_crit = 38 mm (in X7 > X7_crit, RN was predicted). Based on the ROC curve comparison, there were no statistically significant differences between the AUC_{Lin_12} and AUC_{MLP_3} (p = 0.12); thus, the reduced amount of the factor indicators from 12 to 3 did not worsen the model predictive qualities. Designed during primary analysis hypothesis was successfully validated in a referent centre on the cohort of 300 patients. Out of the cohort - 14 (4.6%) patients experienced false positive/negative outcome, which resulted in a radical/partial nephrectomy out of the hypothesis margins. Predictive model is characterized by high sensitivity (95.2%) and specificity (95.4%) in selecting patients for partial nephrectomy. Conclusions: For the polar and lateral tumor locations, the functioning parenchymal volumes of over 58 and 67% respectively serve as PN indications. However, for the medial lesions, the primary PN indication is a tumor size less than 38 mm.