iACP-GCR: Multitask Learning on Graph Convolutional Residual Neural Networks for Screening of Multi-target Anticancer Compounds
T.-H Nguyen-Vo, Trang T. T. Do, and B. P. Nguyen∗
The finding of promising anticancer compounds is vital in anticancer drug development. Recently, besides experimental approaches, several computational advances have been developed to assist experimental scientists to identify potential anticancer candidates. In this study, we proposed a computational framework called iGCR-ACP to predict anticancer compounds. The iGCR-ACP was developed using the NCI-60 dataset, one of the most reliable and well-known sources of experimentally verified compounds. From the dataset, compounds were screened across nine cancer types (panels), including breast, central nervous system, colon, leukemia, non-small cell lung, melanoma, ovarian, prostate, and renal, were collected.
The area under the receiver operating characteristic (ROC) curve (AUCROC) values for all prediction tasks were over 0.95. For the area under the precision-recall (PR) curve (AUCPR) values, the prediction task for the breast panel obtained the smallest AUCPR value of about 0.76 while those for the leukemia and colon panels achieved the greatest AUCPR values of about 0.80. The results showed that the iACP-GCR is a robust and stable framework for determining compounds possessing anticancerous activities. Also, the comparative analysis demonstrated that our approach had achieved better performance compared to state-of-the-art methods.
Source code and data are available upon request.