Modeling protein sequences, structures and functions with deep neural networks

In the rapid-advancing field of biotechnology, proteins - the fundamental building blocks of life - play a critical role in addressing an array of complex biological challenges. As the cost of experimentation is extremely high and various biological datasets have been created, computational methods for understanding proteins have become essential. In this dissertation, we introduced several machine learning algorithms aiming at improving protein structure and function modeling by leveraging data-driven principles. First, we introduce a deep learning approach for protein contact prediction which uses a deep convolutional network to learn meaningful structural motifs based on experimental data. Second, we detail a data-driven method for learning protein structural representation enabling both high-performance and high-efficiency structural searches. Third, we introduce an end-to-end protein network alignment learning algorithm that integrates heterogeneous information from biological network using graph neural networks. In summary, these developments have demonstrated the potential of applying data-driven principle through novel machine learning algorithms to address the challenges in protein modeling, yielding learned biological insights.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2023

Author dc:creator

• Liu, Yang

Contributors dc:contributor

• Peng, Jian
• Han, Jiawei
• Zhai, Chengxiang
• Liu, Qiang

• Machine Learning
• Deep Learning
• Artificial Intelligence
• Computational Biology
• Bioinformatics

Statement dc:rights

• Copyright 2023 Yang Liu

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/121527

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F121527