Modeling Protein Structures andProtein Interactions with DeepLearning Methods

The Hong Kong University of Science and Technology (Guangzhou)

数据科学与分析学域

PhD Thesis Proposal Examination

By Mr. GAO Ziqi

摘要

Protein-protein interactions (PPIs) are essential for the cellular processes of all living organisms. Moreover, accurate PPI prediction has recently been shown to enhance the structure prediction of large-size proteins (i.e., multimers). However, experimental methods for detecting PPIs are often expensive and prone to false positives. Therefore, there is a high demand for efficient computational methods to facilitate PPI detection. With the increasing amount of protein data generated by advanced high throughput technologies, machine learning models have become a promising approach for PPI prediction. In this proposal, we first provide a comprehensive overview of the latest machine learning-based prediction methods for PPIs. Specifically, we comprehensively considered three tasks highly related to PPI: binary PPI prediction, protein-protein docking and multi-PPI prediction. We outline the machine learning models used in these 3 tasks and discuss the protein data representation in detail. To shed light on the future of PPI prediction, we analyze the trends in the development of machine learning-based methods. We also suggest potential directions for PPI prediction, such as incorporating computationally predicted protein structures to expand the data source for machine learning models. Our review aims to serve as a helpful guide for researchers seeking to improve PPI prediction in this field.

TPE Committee

Chair of Committee: Prof. WANG Wei

Prime Supervisor: Prof. LI Jia

Co-Supervisor: Prof. HUANG Yong

Examiner: Prof. DING Zishuo