Traditional drug design methods based on trial and error are costly and inefficient. The computational approach ToSS-MoDE (Topological Substructural Molecular Design) offers an alternative by linking molecular descriptors to biological activity. To develop a QSAR model to predict the anti-inflammatory activity of synthetic and natural compounds using weighted spectral moments. Spectral moments (µk) were calculated from the adjacency matrix between bonds for 410 compounds (180 active and 230 inactive). MODESLAB software (MICROSOFT OFFICE 365) was used to generate descriptors, and Linear Discriminant Analysis (LDA) was applied to classify activity. The model was validated with an external series of 62 compounds. Results. The model showed an overall classification of 91.59% in the training series and 90.2% in validation. The spectral moments µ0, µ3, µ4, and µ5 were the most significant. Diosgenin, a natural metabolite, showed potential anti-inflammatory activity (classification probability: 81%). The model showed strong training performance (91.7% accuracy) and promising external performance for confidently classified compounds. All datasets, descriptor-generation settings, coefficients, and posterior probabilities are fully described in the main text to ensure full reproducibility.
Vueba et al. (2026) studied this question.