Predicting disease-genes based on network information loss and protein complexes in heterogeneous network.

Research on disease-related genes has always been a visible focus in the biological field. This study can help us to reveal the hidden mechanism of human diseases. Although many methods have been developed, the accuracy of identifying disease-genes remains challenging. We propose a computation-based method for predicting disease-genes. The investigations are inspired by the information loss model for evaluating similarities between network nodes and human protein complexes that reflect interactions between genes. Furthermore, we also combine other data such as lncRNA to construct a triple heterogeneous network and design a network propagation algorithm applied to the heterogeneous network (InLPCH). This algorithm effectively reduces the number of false positives in the biological networks when predicting disease-genes and combines the multiple propagation paths of the heterogeneous network to improve prediction accuracy. We conduct extensive experiments over disease-genes dataset. The InLPCH demonstrates high performance in comparison with six other state-of-the-art algorithms.