Microtubules are polymers of alfa and beta tubulin, functioning in many essential cellular processes, including mitosis. Tubulin is characterized by a continue process of polymerization and depolymerization which allow microtubules to stretch and shorten by hydrolysis of GTP to GDP. Chemical compound in the capacity of modify that process, provoke alteration into mitotic spindle that stops cellular proliferation. There are three binding sites on tubulin: taxol, alkaloids of vinca and colchicine. Scope of this work is develop pharmacophoric models at colchicine binding site, evaluating trough application of ROC curves. Models are been designed starting by a training set of 23 ligands available on Protein Data Bank. The evaluation is carried out based on 10 test set compounds with known activity. To identify docking software with best predictive capacity, among Autodock, Gold, Glide Smina and Plants, is carried out a cross docking on 23 selected crystals. The evaluation of ligands is based on comparation with test set of 10 ligand unrelated with known activity and 450 decoys obtained by test set. It was been produced two pharmacophoric models, first one by Phase software, show 5 features: two hydrogen bond acceptors, two aromatic moiety and one hydrophobic region. Second one, manually drive produced, shows 9 features: two aromatic moiety, four hydrophobic region one hydrogen bond donator and two hydrogen bond acceptors. Both pharmacophoric models was utilized to evaluate the test set and decoy database.To define which pharmacophoric model was the best, it was analysed Enrichment Factor (EF), ROC curves and AUC. Pharmacophoric models by Phase has EF of 20% instead manually drive models has EF values of 60%. ROC curves, calculating choose as threshold the fitness value of pharmacophoric models studiedThis allow to value the best predictive capacity and consequently the ability to discern between active compound proposed by manually driven models and Phase models.
Evaluation of pharmacophoric models at colchicine binding site by ROC curves / Esposito, Chiara; Bufano, Marianna; Coluccia, Antonio. - (2019). (Intervento presentato al convegno 6th computational driven drugs discovery, CDDD Meeting 2019 tenutosi a Rome; Italy).
Evaluation of pharmacophoric models at colchicine binding site by ROC curves
Chiara Esposito‡Primo
;Marianna Bufano†Secondo
;Antonio Coluccia†Ultimo
2019
Abstract
Microtubules are polymers of alfa and beta tubulin, functioning in many essential cellular processes, including mitosis. Tubulin is characterized by a continue process of polymerization and depolymerization which allow microtubules to stretch and shorten by hydrolysis of GTP to GDP. Chemical compound in the capacity of modify that process, provoke alteration into mitotic spindle that stops cellular proliferation. There are three binding sites on tubulin: taxol, alkaloids of vinca and colchicine. Scope of this work is develop pharmacophoric models at colchicine binding site, evaluating trough application of ROC curves. Models are been designed starting by a training set of 23 ligands available on Protein Data Bank. The evaluation is carried out based on 10 test set compounds with known activity. To identify docking software with best predictive capacity, among Autodock, Gold, Glide Smina and Plants, is carried out a cross docking on 23 selected crystals. The evaluation of ligands is based on comparation with test set of 10 ligand unrelated with known activity and 450 decoys obtained by test set. It was been produced two pharmacophoric models, first one by Phase software, show 5 features: two hydrogen bond acceptors, two aromatic moiety and one hydrophobic region. Second one, manually drive produced, shows 9 features: two aromatic moiety, four hydrophobic region one hydrogen bond donator and two hydrogen bond acceptors. Both pharmacophoric models was utilized to evaluate the test set and decoy database.To define which pharmacophoric model was the best, it was analysed Enrichment Factor (EF), ROC curves and AUC. Pharmacophoric models by Phase has EF of 20% instead manually drive models has EF values of 60%. ROC curves, calculating choose as threshold the fitness value of pharmacophoric models studiedThis allow to value the best predictive capacity and consequently the ability to discern between active compound proposed by manually driven models and Phase models.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
