Network inference performance complexity: a consequence of topological, experimental and algorithmic determinants

• DOI: 10.1093/bioinformatics/btz105
• Publication Date: 2019-02-14T13:00:40Z
• Authors: Joseph J Muldoon; Jessica S Yu; Mohammad-Kasim Fa...; Neda Bagheri
• Keywords: Blueprint; Gene regulatory network

Network inference algorithms aim to uncover key regulatory interactions governing cellular decision-making, disease progression and therapeutic interventions. Having an accurate blueprint of this regulation is essential for understanding controlling cell behavior. However, the utility impact these approaches are limited because ways in which various factors shape outcomes remain largely unknown.We identify systematically evaluate determinants performance-including network properties, experimental design choices data processing-by developing new metrics that quantify confidence across comparable terms. We conducted a multifactorial analysis demonstrates how stimulus target, kinetics, induction resolution dynamics, noise differentially widely used significant previously unrecognized ways. The results show even if high-quality paired with high-performing algorithms, inferred models sometimes susceptible giving misleading conclusions. Lastly, we validate findings using realistic silico gene networks. This characterization approach provides way more rigorously interpret infer from biological datasets.Code available at http://github.com/bagherilab/networkinference/.Supplementary Bioinformatics online.