Printability of a very-large-scale integration design is assessed by: during a training phase, generating a training set of very-large-scale integration design shapes representative of a population of very-large-scale integration design shapes, obtaining a set of mathematical representations of respective shapes in the training set, identifying at least two classes of physical events causally linked to the printability for the very-large-scale integration design shapes, each of the classes being associated to a respective level of printability, labeling each mathematical representation of the set according to one of the identified classes, based on a lithography model, and selecting a probabilistic model function maximizing a probability of a class, given the set of mathematical representations; and during a testing phase, providing a very-large-scale integration design shape to be tested, testing the provided very-large-scale integration design shape, and labeling the provided very-large-scale integration design shape according to the identified class.