Teaching Development Project: Gene Expression Prediction With Deep Learning
Zhu, Lingyu (2017)
Zhu, Lingyu
2017
Information Technology
Tieto- ja sähkötekniikan tiedekunta - Faculty of Computing and Electrical Engineering
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2017-06-07
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201705241480
https://urn.fi/URN:NBN:fi:tty-201705241480
Tiivistelmä
Histone modifications are playing an important role in affecting gene regulation. In this thesis, a Convolutional Recurrent Neural Network is proposed and applied to predict gene expression levels from histone modification signals. Its two simplified variants: Convolutional Neural Network and Recurrent Neural Network, and one state-of-the-art baseline: DeepChrome are also discussed in this work. Their performances are evaluated with gene expression data that derived from Roadmap Epigenomics Mapping Consortium database by the Receiver Operating Characteristic, Area Under the Curve and statistical analysis. As a result, the Convolutional Recurrent Neural Network model achieves the best performance compared to the other models.
For teaching development of a pattern recognition and machine learning course from Tampere University of Technology, an approach of integrating theory and practice is used. Video recording, weekly exercises and competition are worked as the auxiliary parts of lectures, which helps the students have a better understanding of the theoretical knowledge and learn how to solve different kind of practical problems. We used histone modification data for a competition on this course, and this competition would be discussed emphatically in this thesis. For the competition, we motivated the students to develop multiple machine learning algorithms to accurately predict gene expression levels on five core histone modification masks. During the period of this course, the competition Gene Expression Prediction received 888 entries that submitted by 105 teams with 184 players. This thesis includes the analysis and summary of the outcomes from the competition. Additionally, the learning assessment is also discussed in this thesis.
For teaching development of a pattern recognition and machine learning course from Tampere University of Technology, an approach of integrating theory and practice is used. Video recording, weekly exercises and competition are worked as the auxiliary parts of lectures, which helps the students have a better understanding of the theoretical knowledge and learn how to solve different kind of practical problems. We used histone modification data for a competition on this course, and this competition would be discussed emphatically in this thesis. For the competition, we motivated the students to develop multiple machine learning algorithms to accurately predict gene expression levels on five core histone modification masks. During the period of this course, the competition Gene Expression Prediction received 888 entries that submitted by 105 teams with 184 players. This thesis includes the analysis and summary of the outcomes from the competition. Additionally, the learning assessment is also discussed in this thesis.