Prediction of the protein complex assembly pathway using multiple docking algorithm

Yoichiro Togawa, Purdue University

Abstract

Proteins often function as a complex of multiple subunits, and the quaternary structure is important for proper function. An ordered assembly pathway is one of the strategies nature has developed to obtain the correct conformation: studies have shown a relationship between the assembly pathway and evolution of protein complexes. Identification of the assembly pathway and the intermediate structures helps drug development as well. Therefore, elucidation of the assembly pathway of protein complexes is important for understanding biochemical processes central to cellular function. Recent studies have demonstrated the assembly pathway of a protein complex can be predicted from its crystal structure by comparing the buried surface area (BSA) between each subunit. To our knowledge, this is the first and only work that has predicted the assembly pathways of protein complexes from their structure. In this work, we have developed four methods to predict the assembly pathway from the output of Multi-LZerD, a multiple docking algorithm for asymmetric protein complexes. We found that data from Multi-LZerD predicted not only the model of the complex but also suggested how the complex is assembled. The four methods were benchmarked, along with the BSA-based method, using a dataset of manually-curated protein complexes. In contrast with the data set used in the BSA-based method, which only contained homomeric and symmetric complexes, our data set includes asymmetric complexes varying in size, topology, and number of subunits. We confirmed that the BSA based-method also worked with asymmetric complexes as they predict the correct pathway in 68% of the cases in our data set. Although the success rate of our methods ranges from 40% to 52%, it improved to as high as 82% for the complexes where Multi-LZerD was successful in modeling near native structures. The results also showed that our method is capable of capturing some of the dimerization events in the assembly pathway, even if the overall pathway prediction was failing. Additionally, there was a case where the BSA-based method failed, but our method was successful, suggesting the limitations in the BSA-based method. These results demonstrate the ability of a multiple docking algorithm to predict the assembly pathway of protein complexes.

Degree

M.S.

Advisors

Kihara, Purdue University.

Subject Area

Bioinformatics

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS