Enhancing protein-protein interaction prediction using multiple kernels

• Main Author: Lek, Wei Long
• Other Authors: Kwoh Chee Keong
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Computer science and engineering
• Online Access: http://hdl.handle.net/10356/62609
• Institution: Nanyang Technological University

A protein-protein interaction (PPI) network indicates which pairs of proteins interact. Since proteins hardly perform alone, it is of essential to know which pairs of proteins interact with each other to perform the various bodily functions. However, experimental methods for PPI are tedious, laborious and expensive. Thus, PPI prediction is of interest to researchers as it helps to identify such interactions. For example, functions of unknown or newly discovered proteins may be predicted through similarity with the interactions of similar known protein. Kernel methods have been used to predict PPIs. However, there is always demand for more accuracy. Hence, in this project, we want to enhance the PPI prediction by experimenting with different kernels with the aim of merging the best kernels to obtain improved results. We computed experiments for various kernels and the results were provided in this document.