Predict Molecular Regulatory Network of Norway Rat under the Frame of Data Integration

摘要

Build rat's molecular regulatory network by integrating five heterogeneous data types that serve as evidence for either protein-protein interaction or protein-DNA interaction. P-values for evidence types are calculated by different algorithms and merged together by Support Vector Machines according to estimated weights which indicate respective contributions of different evidence types to the final prediction. Proper classification threshold is specified to effectively control the false discovery rate, and the result is validated by searching predicted interactions in related databases as well as projecting them to signaling pathways to mark up key factors in disease mechanism. An analysis of our methodology versus previous studies and data integration versus single evidence is performed to demonstrate that the solution we present here is more comprehensive and advantageous than traditional ones due to its rational frame structure and full use of information.