Sign assignment problems on protein networks

Shay Houri*, Roded Sharan

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In a maximum sign assignment problem one is given an undirected graph and a set of signed source-target vertex pairs. The goal is to assign signs to the graph's edges so that a maximum number of pairs admit a source-to-target path whose aggregate sign (product of its edge signs) equals the pair's sign. This problem arises in the annotation of physical interaction networks with activation/repression signs. It is known to be NP-complete and most previous approaches to tackle it were limited to considering very short paths in the network. Here we provide a sign assignment algorithm that solves the problem to optimality by reformulating it as an integer program. We apply our algorithm to sign physical interactions in yeast and measure our performance using edges whose activation/repression signs are known. We find that our algorithm achieves high accuracy (89%), outperforming a state-of-the-art method by a significant margin.

Original languageEnglish
Title of host publicationAlgorithms in Bioinformatics - 12th International Workshop, WABI 2012, Proceedings
Pages338-345
Number of pages8
DOIs
StatePublished - 2012
Event12th International Workshop on Algorithms in Bioinformatics, WABI 2012 - Ljubljana, Slovenia
Duration: 10 Sep 201212 Sep 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume7534 LNBI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference12th International Workshop on Algorithms in Bioinformatics, WABI 2012
Country/TerritorySlovenia
CityLjubljana
Period10/09/1212/09/12

Keywords

  • activation
  • integer linear program
  • network annotation
  • protein-protein interaction
  • repression

Fingerprint

Dive into the research topics of 'Sign assignment problems on protein networks'. Together they form a unique fingerprint.

Cite this