Bladder cancer (BLC) is a very dangerous and common disease which is characterized by an uncontrolled growth of
the urinary bladder cells. In the field of chemotherapy, many compounds have been synthesized and evaluated as anti-BLC agents.
The future design of more potent anti-BLC drugs depends on a rigorous and rational discovery, where the computer-aided design
(CADD) methodologies should play a very important role. However, until now, there is no CADD methodology able to predict anti-BLC
activity of compounds versus different BLC cell lines. We report in this work the first unified approach by exploring Quantitative-
Structure Activity Relationship (QSAR) studies using a large and heterogeneous database of compounds. Here, we constructed
two multi-target (mt) QSAR models for the classification of compounds as anti-BLC agents against four BLC cell lines. The first
model was based on linear discriminant analysis (mt-QSAR-LDA) employing fragment-based descriptors while the second model
was obtained using artificial neural networks (mt-QSAR-ANN) with global 2D descriptors. Both models correctly classified more than
90% of active and inactive compounds in training and prediction sets. We also extracted different substructural patterns which could be
responsible for the activity/inactivity of molecules against BLC and we suggested new molecular entities as possible potent and versatile
Keywords: Artificial neural networks, bladder cancer, fragments, in silico design, linear discriminant analysis, molecular descriptors, mt-
QSAR, quantitative contributions.
Rights & PermissionsPrintExport