Please login first
Virtual Screening of 1,2-Diazoles Compounds for Chagas Disease: A Prediction Model
* 1 , 1 , 2 , 1 , 1
1  Post-Graduate Program in Natural and Synthetic Bioactive Products, Federal University of Paraíba, 58051-900 João Pessoa, PB, Brazil;
2  Post-Graduate Program in Natural and Synthetic Bioactive Products, Federal University of Paraíba, 58051-900 João Pessoa, PB, Brazil; Center of Education and Health, Federal University of Campina Grande. 58175-000 Cuité, PB, Brazil.


INTRODUCTION: Chagas' disease is a parasitic, chronic and emergency infection caused by the protozoan Trypanosoma cruzi[1]. Because of this, we are constantly seeking new therapeutic alternatives and methods of study. OBJECTIVES: To perform a virtual screening of 1,2-diazoles compounds as potential therapeutic agents for Chagas' disease through the elaboration of a predictive model. METHODS: A CHEMBL database[2], composed of 661 chemical structures with activity potential against Trypanosoma cruzi, was selected. The compounds were classified according to the pIC50 value (-log of IC50), with 332 active (pIC50 ≥7.1) and 328 inactive (pIC50 <7.1) that was selected randomly, keeping the same ratio of active/inactive compounds, in training set composed of 528 active and 129 inactive compounds and test set with 133 active and 129 inactive samples. The prediction set is composed of 31 unpublished 1,2-diazole compounds. The SMILES codes were the input data for all structures, which were in the Standardizer and predicted their properties generated by the VolSurf software[3]. The model was generated by KNIME 3.1.0 software, using the Random Forest (RF) calculation algorithm. RESULTS AND DISCUSSION: In the analysis of the model, the hit rates obtained in the test and cross-validation were higher than 73%. In both, this parameter for active compounds was higher than the inactive ones, being, respectively, 80% and 80% for the test and 78% and 73% for the cross validation. The classification rate of the model was evaluated by the graph Receiver Operating Characteristic, corresponding in the test set to 0.843, indicating a high classification rate. The Matthews Correlation Coefficient was used to evaluate the prediction of the model, resulting in 0.63 in the test and 0.61 in the cross validation, indicating that the model has a good prediction. The RF model demonstrated that 13 molecules studied showed a percentage of potential activity above 65%. CONCLUSION: The model presented accuracy, reproducibility and distinguished the probability of potential activity of the molecules under study.


  1. WORLD HEALTH ORGANIZATION, Chagas Disease (American Trypanosomiasis). Geneva: 2018. Disponível em: Acesso em: 10 de Julho de 2018.
  2. EUROPEAN BIOINFORMATICS INSTITUTE. CHEMBL Database. Available in: Access in: 10 july of 2018.
  3. MOLECULAR DISCOVERY. Vol Surf (Volume and Surface Descriptors). Available in: Access in: 10 july of 2018.
Keywords: Keywords: Diazoles; Prediction; Virtual Screening; Random Forest.