Colorectal cancer (CRC) appears to be the third most common cancer as well as the fourth most common cause of cancer deaths in the world. Its most lethal states are when it becomes metastatic. It is of interest to find tests that can quickly and accurately determine if the patient has already developed metastasis. Changes in methylation profiles have been found to be characteristic of cancers at different stages and can therefore be used to develop diagnostic panels. We developed a deep learning (DL) model (Deep2Met) using methylation profiles of patients with CRC to predict if the cancer is in its metastatic state. Results suggest that our method achieves an AUPR and an average F-score of 96.99% and 94.71%, respectively, making Deep2Met potentially useful for diagnostic purposes. The DL model Deep2Met we developed, shows promise in the diagnosis of CRC based on methylation profiles of individual patients.
|Original language||English (US)|
|Title of host publication||Proceedings of the 2019 6th International Conference on Bioinformatics Research and Applications|
|Number of pages||6|
|State||Published - May 4 2020|
Bibliographical noteKAUST Repository Item: Exported on 2020-10-01
Acknowledged KAUST grant number(s): BAS/1/1606-01-01, URF/1/1976
Acknowledgements: This work has been supported by the King Abdullah University of Science and Technology (KAUST) Base Research Fund (BAS/1/1606-01-01) to VBB, and KAUST Office of Sponsored Research (OSR) under Awards No CCF ? URF/1/1976-30-01