An Integrative Multi-Network and Multi-Classifier Approach to Predict Genetic Interactions

摘要: Genetic interactions occur when a combination of mutations results in surprising phenotype. These capture functional redundancy, and thus are important for predicting function, dissecting protein complexes into pathways, exploring the mechanistic underpinnings common human diseases. Synthetic sickness lethality most studied types genetic yeast. However, even yeast, only small proportion gene pairs have been tested due to large number possible combinations pairs. To expand set known synthetic lethal (SL) interactions, we devised an integrative, multi-network approach these that significantly improves upon existing approaches. First, defined features characterizing relationships between genes from various data sources. In particular, independent SL contrast some previous Using features, developed non-parametric multi-classifier system enabled simultaneous use multiple classification procedures. Several comprehensive experiments demonstrated SL-independent conjunction with advanced scheme led improved performance compared current state art method. this approach, derived first yeast transcription factor interaction network, part which was well supported by literature. We also used predict all non-essential (http://sage.fhcrc.org/downloads/downloads/predicted_yeast_genetic_interactions.zip). This integrative is expected be more effective robust uncovering new tens millions unknown hundreds higher organisms like mouse human, very few identified date.