Plasmodium falciparum glucose-6-phosphate dehydrogenase (PfG6PD) has been taken as a
promising target for developing anti-malarial drugs. In this work, four quantitative models were generated
to predict the bioactivity of 262 Pf G6PD inhibitors. Each molecule was represented by ADRIANA.
Code descriptors. The whole dataset was divided into a training set and a test set randomly or by
a Kohonen’s self-organizing map (SOM). Afterwards, the bioactivities of Pf G6PD inhibitors were
predicted with Multiple Linear Regression (MLR) and Support Vector Machine (SVM) method, respectively.
All the four models gave a correlation coefficient (r) over 0.82 and a RMSE under 0.18 on test set. By analyzing
the descriptors in these models, it was observed that number of hydrogen bonding acceptors, ring complexity, σ atom
charges, total atom charges and effective atom polarizabilitly properties are important for the bioactivity of Pf G6PD inhibitors.
Keywords: Plasmodium falciparum glucose-6-phosphate dehydrogenase (PfG6PD) inhibitor, kohonen’s self-organizing map
(SOM), multiple linear regression (MLR), support vector machine (SVM), quantitative structure-activity relationship
Rights & PermissionsPrintExport