FACULDADE DE ENGENHARIA DA UNIVERSIDADE DO PORTO Application of Machine Learning techniques on the Discovery and annotation of Transposons in genomes

Transposable elements or transposons are sequences of DNA that move and transpose within a genome. Known as mutation agents, these elements are broadly studied due to their importance in disease research, genome alteration and because of their importance on species evolution. Several methods were developed to discover and annotate transposable elements and they are classified in four main categories: De novo, Structure-based, Comparative Genomic and Homology-based. There are different tools based on these methodologies that detect transposable elements, al-though, there isn’t any single tool which has good results in detecting all the different types of transposable elements. Taking this into account, this dissertation will have three distinct phases. At first there will be generated datasets of curated DNA sequences with transposable elements inserted in known positions. These datasets will be as diversified as possible so they can cover all the different scenarios found in real genomes. Following, using these datasets, transposon detection tools are evaluated and their accuracy is measured. Using the results of the transposon detection tools ’ evaluation, the last step of this dissertation is to use Machine Learning techniques to identify and characterize the context where these tools fail short in transposon detection. After this, the aim is to create a meta-tool using models gen-erated by Machine Learning that combines the best of different transposable elements detection tools in order to improve the overall detection accuracy. i i