Russian researchers from St. Petersburg University and Samara State Technical University have developed a method for analyzing crystal structure using neural networks. The approach combines topological assessment of crystals with artificial intelligence tools: a special program determines the proximity of atoms based on the geometric structure of the material, and a neural network reveals statistical patterns by analyzing possible combinations of contact between atoms. This allows you to find connections that are hard to find in a normal search. Using artificial intelligence, scientists have identified a number of previously unknown interactions between halogen atoms and other elements, the nature of which has not yet been explored by quantum chemical calculations.