RCPred: RNA complex prediction as a constrained maximum weight clique problem
Résumé
Background: RNAs can interact and form complexes, which have various biological roles. The secondary structureprediction of those complexes is a first step towards the identification of their 3D structure. We propose an originalapproach that takes advantage of the high number of RNA secondary structure and RNA-RNA interaction predictiontools. We formulate the problem of RNA complex prediction as the determination of the best combination (accordingto the free energy) of predicted RNA secondary structures and RNA-RNA interactions.Results: We model those predicted structures and interactions as a graph in order to have a combinatorialoptimization problem that is a constrained maximum weight clique problem. We propose an heuristic based onBreakout Local Search to solve this problem and a tool, called RCPred, that returns several solutions, including motifslike internal and external pseudoknots. On a large number of complexes, RCPred gives competitive results comparedto the methods of the state of the art.Conclusions: We propose in this paper a method called RCPred for the prediction of several secondary structures ofRNA complexes, including internal and external pseudoknots. As further works we will propose an improvedcomputation of the global energy and the insertion of 3D motifs in the RNA complexes.