ENNET: inferring large gene regulatory networks from expression data using gradient boosting

Abstract

BackgroundThe regulation of gene expression by transcription factors is a key determinant of cellular phenotypes. Deciphering genome-wide networks that capture which transcription factors regulate which genes is one of the major efforts towards understanding and accurate modeling of living systems. However, reverse-engineering the network from gene expression profiles remains a challenge, because the data are noisy, high dimensional and sparse, and the regulation is often obscured by indirect connections.ResultsWe introduce a gene regulatory network inference algorithm ENNET, which reverse-engineers networks of transcriptional regulation from a variety of expression profiles with a superior accuracy compared to the state-of-the-art methods. The proposed method relies on the boosting of regression stumps combined with a relative variable importance measure for the initial scoring of transcription factors with respect to each gene. Then, we propose a technique for using a distribution of the initial scores and information about knockouts to refine the predictions. We evaluated the proposed method on the DREAM3, DREAM4 and DREAM5 data sets and achieved higher accuracy than the winners of those competitions and other established methods.ConclusionsSuperior accuracy achieved on the three different benchmark data sets shows that ENNET is a top contender in the task of network inference. It is a versatile method that uses information about which gene was knocked-out in which experiment if it is available, but remains the top performer even without such information. ENNET is available for download from https://github.com/slawekj/ennet under the GNU GPLv3 license.

BTNET : boosted tree based gene regulatory network inference algorithm using time-course measurement data

“…Gradient boosting was also successfully used for inferring gene regulatory networks from steady-state gene expression data [9, 30]. The gradient boosting algorithm follows the gradient descent procedure that is employed to minimize the loss L of an estimator f by adding residual fitted estimator h [28].…”

“…A good deal of research on reverse-engineering has been conducted using the gene expression data [6–9]. In the DREAM (Dialogue for Reverse Engineering Assessments and Methods) Challenges, methods were employed to construct a benchmark dataset that can be used to validate various inference algorithms [10, 11].…”

Gene Networks in Plant Biology: Approaches in Reconstruction and Analysis

“…Boosting utilizes an iterative strategy to slowly improve the model and should be highly accurate. For the DREAM5 datasets, the community network based on meta-analysis was robust and outperformed individual methods, and two ensemble-based methods achieved higher accuracy than individual methods [47,63].…”

“…For example, correlation coefficient methods are more reliable than regression methods for feed-forward loops, whereas regression methods perform better for linear cascades. Because of the complementarity of various methods, metaprediction approaches that combine multiple methods may be able to outperform individual methods [47,50,[61][62][63].…”

Gene regulatory network inference using PLS-based methods

“…As using random forests for feature selection is not well understood theoretically, the TIGRESS method [20] uses least angle regression (LARS) with stability selection combined to solve the GRN inference problem. The ENNET method [21] aggregates the features selected by an algorithm based on Gradient Boosting Machine. However, the ENNET method has high computational cost when it is applied on the high-dimensional data (i.e., the data with thousands of features).…”