ChemPick is dedicated to visualizing and compare chemical inventories in multidimensional descriptor space including 2D and 3D molecular parameters. The system could categorize chemicals based on their reactivity, solubility, toxicity, etc., using mechanistically relevant descriptors. Probabilistic assessment of model applicability domain and domains with high biological activity is provided. One of the most important features of the system is capability to select chemicals for strategic testing. The chemical selection from an inventory could be oriented either to expend the structural domain of a training set or to QSAR model validation purpose. Two groups of criteria are currently used for chemicals selecting. First of them is based on Euclidean metrics provides information about training set interpolation and extrapolation area as well as the furthest reaches from the training set. One can select chemicals evenly distributed in the interested area and control their number by changing the granularity of the lattice. Testing of chemicals from training set extrapolation area could effectively expend the structural diversity of training set and derive more general QSAR models. Similarly, inventory chemicals from training set interpolation area could be tested for identifying potentially active chemicals. The second group of methods for selecting chemicals is based on probabilistic models derived from the training set. They allow selection of inventory chemicals which meet certain probability thresholds to be active and/or to belong to model applicability domain with a specified density population.