Cannabinoid (CB) ligands have potential therapeutic applications in treating many immune system disorders and accurate predictions of their CB receptor binding activities play an important role in finding potent CB ligands. Structural diversity of CB ligands presents a challenging task to develop a model which can generalize across different chemical scaffolds. In this work, we developed a molecular fingerprint-based artificial neural network QSAR approach (ANN-QSAR) to predict binding activities of structurally diverse CB ligands. A total of 1699 reported CB2 ligands and their binding affinity data (Ki) were used. Different fingerprints such as MACCS and ECFP4 of CB ligands were used as inputs to train the ANN-QSAR model. Upon successful model training and validation, the prediction accuracy (r2 value of test set) of ANN-QSAR model was 0.62 which proved that a combination of molecular fingerprints and ANN method provides an efficient route for SAR studies of structurally diverse chemical entities.

Artificial neural network approach to predict GPCR cannabinoid ligand binding activities

243rd National Meeting (2012)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.