Deep learning for Direct DNA Domain Detection
Master thesis
Permanent lenke
https://hdl.handle.net/11250/3147983Utgivelsesdato
2024Metadata
Vis full innførselSamlinger
- Master's theses (RealTek) [1826]
Sammendrag
As more and more genomes are sequences it becomes ever more needed to have access to fastmethods of analyzing them. One such analysis is finding regions of the genome that containdescriptions of proteins we are familiar with. Pfam is a database containing these descriptions,and tools like HMMER can match known proteins/protein substructures onto sequences. Thisprocess is slow, and this thesis explores how to speed it up by creating machine-learning modelsthat filter out the sequences less likely to contain known proteins.This is done by analyzing large datasets containing hundreds of species sampled by differentmeans and by employing different target values. Through this valuable insight such as highGC content negatively impacting model performance and that machine learning models suchas convolutional neural networks can describe specific Pfam patterns close to perfectly, withaccuracies of 99.7% on balanced datasets.The end result of the thesis is that a filtering mechanism should be possible to create, but itwould require significantly more work to get it working to a degree where it both included thenecessary. Ensuring that enough Pfam entries are represented would be a good starting step.One of the models this thesis employs is the Tsetlin machine, hoping its unique structure wouldmake it well suited to such data. The results did, however, not show this to be the case. Whilethe Tsetlin machine might still be well suited for similar data, some changes would have to bemade to how they were employed in this thesis.