Metal-organic frameworks (MOFs) attract a lot of interest due to their unique structure-dependent properties. Their internal pores comparable to the size of small molecules are naturally refined for various absorbance effects. Possessed properties lie in a foundation of multiple applications, such as catalysis, gas storage/separation and especially – clean energy related ones. Theoretical calculations are a usual way of decreasing experimental costs while investigating properties of new materials, especially at a design stage. Electronic structure calculations like density functional theory (DFT) in most cases provide an appropriate accuracy in matching experimentally measured data such as adsorbate interaction energies. However, as in the case of experimental studies, large-scale materials screening studies with DFT calculations are rather time-consuming, and it can be carried out only for structures with relatively small unit cell. Here we would like to present a theoretical and experimental results describing calculation of electron density in metal-organic frameworks. We built a model trained to predict partial charges on MOF atoms based on DFT calculations. The relative error of the model allows us to conclude that models do not decrease the level of accuracy and do not superinduce additional error comparing to DFT. At the same time, computational cost of the model is several orders of magnitude less. Models also demonstrated transferability and allowed to make prediction e.g. for MOFs containing metals not presented in the train set. We have also built a force-field (FF) of two-centered and three-centered interatomic potentials constructed using predicted charges. The FF proved to reproduce MOF crystal structure. As a final test, we have applied the developed model and FF to a new synthesized lanthanide-containing MOFs to estimate influence of supramolecular effects on metal complexation selectivity. As a result, we’ve built a model predicting one of basic MOF properties within relatively small computational time and tested it on experimental data, both obtained from literature sources and self-investigated.