Prior Knowledge Guided Gene-Disease Associations Prediction: An Enhanced Inductive Matrix Completion Approach

Abstract

Exploring gene-disease associations is of great significance for early prevention, diagnosis and treatment of diseases. Most existing methods depend on specific type of biological evidence and thus are limited in the application. More importantly, these methods ignore some inherent prior sparsity and structure knowledge which is useful for predicting gene-disease associations. To address these challenges, a novel Enhanced Inductive Matrix Completion (EIMC) model is proposed to predict pathogenic genes by introducing the prior sparsity and structure knowledge into the traditional Inductive Matrix Completion (IMC). Specifically, the EIMC model not only employs the sparse regularization to preserve the prior sparsity of gene-disease associations, but also employs the manifold regularization to capture the prior structure information of data distribution. To the best of our knowledge, the proposed EIMC is the first model to simultaneously incorporate both prior sparse and manifold regularizations into the same objective function. Additionally, note that our proposed EIMC model also integrates the features of genes and diseases extracted from various types of biological data, and can predict new genes and diseases by using an inductive learning strategy. Finally, the extensive experimental results demonstrate that our proposed model outperforms other state-of-the-art methods.