Generating full-scale libraries from small-scale libraries through the use of generative adversarial networks

• Main Author: Sim, Ryan Yao Rong
• Other Authors: Melissa Jane Fullwood
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Medicine, Health and Life Sciences; Generative adversarial networks; RNA-seq
• Online Access: https://hdl.handle.net/10356/175408
• Institution: Nanyang Technological University

RNA sequencing (RNA-seq) is a useful tool for investigating gene expression and transcriptomic landscapes in various biological contexts. (Wang. Z et al.,2010) However, conducting large-scale RNA-seq experiments can be expensive, resource-intensive, and difficult to obtain as sometimes there are legal and ethical problems involved if human-identifiable data is involved. (Ai, X.,2023) To overcome this issue, researchers have explored the possibility of creating large-scale RNA-seq libraries from smaller datasets using Generative Adversarial Networks (GANs). The aim is to determine whether small-scale libraries are able to encapsulate the core elements of large-scale RNA-seq libraries and to determine whether GANs are able to generate synthetic data that mimics the original dataset and preserves the biological correlation between the samples in the datasets.