Uso de métodos in silico para a identificação de alcaloides com atividade em múltiplos alvos no tratamento da doença de Alzheimer

Abstract

Alzheimer's disease is a prevalent and progressive neurodegenerative disease and the leading cause of dementia worldwide. Despite its widespread incidence, available treatments are scarce. Currently, the search for new therapeutic options has focused on the cholinergic and glutamatergic pathways. In silico approaches are essential to identify promising substances, reducing research costs and time. The aim of this study is to identify alkaloids with potential activity through virtual screening using QSAR and docking models. QSAR models were developed from banks of structures tested on these targets. 2D and 3D descriptors were calculated and evaluated for multicollinearity and correlation. Models were developed with different algorithms, undergoing internal and external validation, and used to screen alkaloid banks. The NMDA, NAChR and AChE target banks contain 91, 100 and 389 chemical structures, respectively, for which 7974 descriptors were calculated. The crystallographic structures coded 5EWM, 3SQ6 and 4EY7 were used for docking, and the models developed were used to evaluate the behavior of the alkaloids selected for the respective sites and the interactions carried out. For the NMDA model, 8 descriptors were selected, with the best result being the MLP model with R2TEST=0.654 and Q2=0.533. For the NAChR model, 6 descriptors were used, with the MLP model showing R2TESTE=0.861 and Q2=0.672. Finally, for the AChE model, 7 descriptors were selected, with the best MLP model showing R2TEST=0.872 and Q2=0.605. It was observed that in all the models developed, one or more descriptors were selected that emphasized the importance of nitrogen compounds for target activity. This characteristic is observed in endogenous ligands and drugs known to act on them. After validation, the models were used to screen the alkaloid bank, evaluating compounds with the highest predicted activity according to the target's activity in the disease. It was noted that the predicted activity values were in line with the experimental values of active substances in the targets. In addition, many of these substances have reported neuroprotective activity, or belong to classes of alkaloids commonly associated with this activity. Substances already tested on the target were also identified, such as boldine selected for AChE. The results of the screening were submitted to docking models, and those with the best performance and favorable interactions with the amino acids that make up the active site were selected. This study demonstrated the efficiency of using in silico methods to identify alkaloids with potential activity against the targets addressed. The compounds identified represent important options for the development of new therapeutic options for Alzheimer's disease.