Deep Learning for genomic data analysis

• Autor(a) principal: Vítor Filipe Oliveira Teixeira
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://repositorio-aberto.up.pt/handle/10216/106492
• topic: Engenharia electrotécnica, electrónica e informática; Electrical engineering, Electronic engineering, Information engineering

Since the Human Genome Project, the availability of genomic data has largely increased. In the last years, genome sequencing technologies and techniques have been improving at a fast rate, resulting in a cheaper and faster genome sequencing. Such amount of data enables both more complex analysis and advances in research. However, a sequencing process quite often produces a huge amount of data that is highly complex. A considerable computational power and efficient algorithms are mandatory in order to extract useful information and perform it in reasonable time, which can represent a constraint on the extraction and comprehension of such information.In this work, we focus on the biological aspects of RNA-Seq and its analysis using traditional Machine Learning and Deep learning methods. We divided our study into two branches. First, we built and compared the accuracy of classifiers that were able distinguish the RNA-seq samples of thyroid cancer patients from samples of healthy persons. Secondly, we have investigated the possibility of building comprehensible descriptions for the differences in the RNA-Seq data by using Denoising Autoencoders and Stacked Denoising Autoencoders as base classifiers and then devising post-processing techniques to extract comprehensible and biologically meaningful descriptions out of the constructed models.