Topological Evaluation of Methods for Reconstruction of Genetic Regulatory Networks

Network inference is advancing rapidly, and new methods are proposed on a regular basis. Understanding the advantages and limitations of different network inference methods is key to their effective application in different circumstances. The common structural properties shared by diverse networks naturally pose a challenge when it comes to devising accurate inference methods, but surprisingly, there is a paucity of comparison and evaluation methods. Historically, every new methodology has only been tested against "gold standard" purpose-designed synthetic and real-world (validated) biological networks. In this paper we aim to assess the impact of taking into consideration topological aspects on the evaluation of the final accuracy of an inference procedure. Specifically, we will compare the best inference methods, in statistical terms, for preserving the topological properties of synthetic and biological networks. A new method for performance comparison is introduced by borrowing ideas from gene set enrichment analysis. Experimental results show that no individual algorithm stands out among the three inference tasks assessed, and the challenging nature of network inference is evident in the struggle of some of the algorithms to turn in a performance that's better than random guesswork. Therefore care should be taken to suit the method used to the specific purpose.