Deep learning Approach for identifying signatures for Ovarian Serous Cystadenocarcinoma

Universiti Kebangsaan Malaysia | Malaysia

Sin Min Tan | Jee Yuan Goh | Abdul Arriff Abdul Jalal | Izzura Malindo

Ovarian cancer is a complex disease which has yet been fully understood. Thus, the diagnosis and treatment of ovarian cancer remain poor. This has led to the poor prognosis of ovarian cancer where only 40% of patients survive 5 years. Understanding of ovarian cancer biology can be accelerated using computational methods to analyze large ovarian cancer dataset. The Cancer Genome Atlas database (TCGA) provides over 30 terabytes of ovarian cancer data including genotyping arrays, methylation arrays, whole exome sequencing RNA-seq and miRNA-seq. Ovarian Serous Cystadenocarcinoma is a type of epithelial ovarian cancer and accounts for about 90 percent of all ovarian cancers. This project aims to identify biological features of ovarian cancer survivors. We have three smaller aims to answer the main aim: to identify features in ovarian cancer survivors that are not found in non-survivors and to characterize identified features in relation to ovarian cancer biology. To achieve the aims, we will do data mining from the TCGA database. The next phase will output the deep learning model which can differentiate omics data between survivors and non-survivors of ovarian cancer. The final phase will characterize the biological features and put a confidence scoring onto the features. Outputs of this project include a list of identified biological features in ovarian cancer survivors. These features may then be further developed into diagnostics methods for ovarian cancer. They will also help in elucidating ovarian cancer biology. We believe this project is essential as part of a long-term translational research where the understanding of ovarian cancer biology leads to applications in medical practices and precision medicine.