TY - GEN
T1 - Protein interaction inference as a Max-SAT problem
AU - Zhang, Ya
AU - Zha, Hongyuan
AU - Chu, Chao Hisen
AU - Ji, Xiang
N1 - Publisher Copyright:
© 2005 IEEE Computer Society. All rights reserved.
PY - 2005
Y1 - 2005
N2 - Discovering interacting proteins is essential for understanding protein functions. However, high throughput interaction data are inherently noisy and only cover a small portion of the whole interactome. Domains, the building block of proteins, are believed to be responsible for the interactions among proteins. An abstract representation of interactome is achieved at domain level and this representation also facilitates the discovery of unobserved protein-protein interactions. Many domain-based approaches have been proposed to predict protein-protein interactions and promising results have been obtained. Existing methods generally assume that domain interactions are independent of each other for the convenience of computational modeling. In this paper, a new framework of learning is proposed. The framework makes no assumption about domain interactions and consider protein interactions resulting from multiple domain interactions which may be dependent of each other. With a conjunctive normal form representation of the relationship between protein interactions and domain interactions, the problem of interaction inference is modeled as a constraint satisfiability problem and solved via linear programming. Experimental results on a combined yeast data set have demonstrated the robustness of and the accuracy of the proposed algorithm.
AB - Discovering interacting proteins is essential for understanding protein functions. However, high throughput interaction data are inherently noisy and only cover a small portion of the whole interactome. Domains, the building block of proteins, are believed to be responsible for the interactions among proteins. An abstract representation of interactome is achieved at domain level and this representation also facilitates the discovery of unobserved protein-protein interactions. Many domain-based approaches have been proposed to predict protein-protein interactions and promising results have been obtained. Existing methods generally assume that domain interactions are independent of each other for the convenience of computational modeling. In this paper, a new framework of learning is proposed. The framework makes no assumption about domain interactions and consider protein interactions resulting from multiple domain interactions which may be dependent of each other. With a conjunctive normal form representation of the relationship between protein interactions and domain interactions, the problem of interaction inference is modeled as a constraint satisfiability problem and solved via linear programming. Experimental results on a combined yeast data set have demonstrated the robustness of and the accuracy of the proposed algorithm.
UR - http://www.scopus.com/inward/record.url?scp=84923032608&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84923032608&partnerID=8YFLogxK
U2 - 10.1109/CVPR.2005.515
DO - 10.1109/CVPR.2005.515
M3 - Conference contribution
AN - SCOPUS:84923032608
T3 - IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops
BT - 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, CVPR 2005 - Workshops
PB - IEEE Computer Society
T2 - 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, CVPR 2005 - Workshops
Y2 - 21 September 2005 through 23 September 2005
ER -