NEArender: an R package for functional interpretation of ‘omics’ data via network enrichment analysis

BMC Bioinformatics

Research

Ashwini Jeggari¹  Andrey Alexeyenko² 

[1] Department of Cell and Molecular Biology, Karolinska Institutet, Stockholm, Sweden;National Bioinformatics Infrastructure Sweden, Science for Life Laboratory, Stockholm, Sweden;Department of Microbiology, Tumor and Cell biology, Karolinska Institutet, Stockholm, Sweden;

关键词: Enrichment;    Network analysis;    Network benchmark;    R package;   

DOI  :  10.1186/s12859-017-1534-y

来源: Springer

【 摘 要 】

BackgroundThe statistical evaluation of pathway enrichment, i.e. of gene profiles' confluence to the pathway level, allows exploring molecular landscapes using functionally annotated gene sets. However, pathway scores can also be used as predictive features in machine learning. That requires, firstly, increasing statistical power and biological relevance via a network enrichment analysis (NEA) and, secondly, a fast and convenient procedure for rendering the original data into a space of pathway scores. However, previous implementations of NEA involved multiple runs of network randomization and were therefore slow.ResultsHere, we present a new R package NEArender which can transform raw 'omics' features of experimental or clinical samples into matrices describing the same samples with many fewer NEA-based pathway scores. This is done via a parametric estimation of the null binomial distribution and is thus much faster and less biased than randomization procedures. Further, we compare estimates from these two alternative procedures and demonstrate that the summarization of individual genes to pathways increases the statistical power compared to both the default differential expression analysis on individual genes and the state-of-the-art gene set enrichment analysis. The package also contains functions for preparing input, modeling null distributions, and evaluating alternative versions of the global network.ConclusionsBeyond the state-of-the-art exploration of molecular data through pathway enrichment, score matrices produced by NEArender can be used in larger bioinformatics pipelines as input for phenotype modeling, predicting disease outcomes etc. This approach is often more sensitive and robust than using the original data. The package NEArender is complementary to the online NEA tool EviNet (https://www.evinet.org) and, unlike of the latter, enables high performance of computations off-line.The R package NEArender version 1.4 is available at CRAN repositoryhttps://cran.r-project.org/web/packages/NEArender/

【 授权许可】

CC BY   
© The Author(s). 2017

【 图 表 】

12951_2017_255_Article_IEq50.gif

12951_2017_255_Article_IEq49.gif

12936_2017_1904_Article_IEq1.gif

Fig. 1

Fig. 5

Fig. 6