Due to the high interest that represents the study from the adverse reactions to the medications and the importance of foreseeing the same ones the employment of the methods of molecular modeling with such an becomes a novel fact. In this work was carried out the calculation of the spectral moments of the adjacency matrix between edges of the molecular graph with suppressed hydrogens, pondered in the main diagonal with different parameters that characterize as much to the connections as to the atoms in the molecules of 63 compound of analgesic action, using for it the methodology MODESLAB. 91 descriptors were calculated, which were used in a series of training divided in four groups, according to the type of adverse reaction but it frequents. With the objective of identifying the descriptors that better they discriminant against among those compound of each group and to define the group of functions of these descriptors able to distinguish with the biggest precision possible to the members of one or another group, an discriminant analysis was developed by means of the statistical software Statistical 8.5. three functions were generated that they constitute lineal combinations of 6 molecular descriptors, which code so much information steric as electronic of the molecules of each group. The obtained functions present a Lambda of very low minimum Wilks (0,07) and a high canonical correlation (0,82), that which demonstrates their power discriminant