XGRN: Reconstruction of Biological Networks Based on Boosted Trees Regression

In Systems Biology, the complex relationships between different entities in cells are modeled and analyzed using networks. Towards this aim, a rich variety of gene regulatory network (GRN) inference algorithms has been developed recent years. However, most rely solely on expression data to reconstruct network. Due possible profile similarity, predictions can contain connections biologically unrelated genes. Therefore, previously known biological information should also be considered by computational methods obtain more consistent results, such as experimentally validated interactions transcription factors target work, we propose XGBoost for networks (XGRN), supervised algorithm, which combines with GRN inference. The key idea our method is train regression model each interaction then utilize predict new interactions. performed XGBoost, state-of-the-art algorithm an ensemble decision trees. detail, XGRN learns based two interactors provides input other candidate interactors. Application benchmark datasets real large single-cell RNA-Seq experiment resulted high performance compared unsupervised methods, demonstrating ability provide reliable predictions.