Prediction of novel angiogenesis inhibitors using in silico method

Angiogenesis plays in important role in tissue repair and could cause all kind of complications when its balance is disturbed. Cancer and angiogenesis is closely related and inhibiting angiogenesis in cancer is heavily studied. Humanized monoclonal antibody such as bevacizumab is one of many angiogenesis inhibitors available in the market. However, it is administered intravenously and the treatment can be very expensive. The purpose of this study is to build a computational model that first analyzes protein-ligand binding patterns of anti-angiogenesis drugs for the purpose of predicting a novel angiogenesis inhibitor. 12 different angiogenesis receptors studied and compounds associated with them were obtained from ChEMBL database and serves as the training set. 8 different prediction models were built, which were from the combination of different fingerprints (ECFP_4, FCFP_4, PubChem, MACCS) and machine learning algorithm (Naive Bayes, Decision Tree). The combination of MACCS-Decision Tree performed the best, with a sensitivity and specificity values of 0.92 at rank 5. The MACCS­ Decision Tree model was then subjected to external validation where 4 compounds; Shiraiachrome-A, 11, 11'-dideoxyverticilin, Quercetin and TKI-31, obtained from scientific literature were tested. The model was able to predict the correct target for 3 of the compounds. This goes to show that the model can be used to discover novel anti-angiogenesis drugs. Future work should include the in vivo or in vitro validation of the in silico result.