COMPUTATIONAL APPROACHES FOR OVARIAN CANCER DRUG REPURPOSING AND GUT MICROBIOME-BASED COLORECTAL DISEASE CLASSIFICATION

Abstract

The thesis focuses on the computational approaches for diagnosis of disease and drug discovery. The work combines classification of colorectal disease based on gut microbiome using machine learning and transcriptomics drug repurposing for the treatment of ovarian cancer. In the colorectal disease study, the analysis of gut microbiome data which was collected from stool sample is done to identify differences between healthy individuals and colorectal disease patients. There are various bioinformatics tools such as SRA Toolkit, Kraken2, Bowtie2 and Bracken which has been used to process the publicly available metagenomics datasets. Preprocessing is followed by CLR normalization and disease classification with the help of different machine learning models such as Logistic Regression, SVM, Random Forest, XGBoost, LightGBM, and KNN. The best performance with high accuracy and ROC-AUC score is shown by Logistic Regression combined with SVM(RBF). With the help of results, we can say that gut microbiome profiles may serve as biomarkers for colorectal disease detection. In the ovarian cancer study, we applied approaches for drug repurposing for identification of potential EPCAM inhibitor molecules.. Drug prediction and screening is done by using tools such as LINCS L1000, SwissADME. Further procedure is carried out for molecular docking analysis using AutoDock Vina. There were several compounds which was shortlisted, AT-7519 showed the best binding affinity against the EPCAM protein and lead to indication that AT-7519 has the potential for the treatment of ovarian cancer and can be used as repurposed drug candidate. Overall, the findings of the studies suggest the importance of gut microbiome profiles which can act as a potential biomarker for detection of colorectal disease. Thus, the main highlights of the thesis are use of machine learning, metagenomics, transcriptomics and molecular docking for in-silico drug discovery and precision medicine.