Gene set enrichment for reproducible science: comparison of CERNO and eight other algorithms

• DOI: 10.1093/bioinformatics/btz447
• Publication Date: 2019-05-31T19:12:17Z
• Authors: Joanna Zyla; Michal Marczyk; Teresa Domaszewska; Stefan H E Kaufmann; Joanna Polanska; January Weiner
• Keywords: R package

Analysis of gene set (GS) enrichment is an essential part functional omics studies. Here, we complement the established evaluation metrics GS algorithms with a novel approach to assess practical reproducibility scientific results obtained from tests when applied related data different studies.We evaluated eight and one algorithm for reproducibility, sensitivity, prioritization, false positive rate computational time. In addition algorithms, also included Coincident Extreme Ranks in Numerical Observations (CERNO), flexible fast based on modified Fisher P-value integration. Using real-world datasets, demonstrate that CERNO robust ranking metrics, as well sample size. had highest while remaining sensitive, specific fast. overall Pathway Down-weighting Overlapping Genes, over-representation analysis performed best, GeneSetTest scored high terms reproducibility.tmod package implementing available CRAN (cran.r-project.org/web/packages/tmod/index.html) online implementation can be found at http://tmod.online/. The datasets analyzed this study are widely KEGGdzPathwaysGEO, KEGGandMetacoreDzPathwaysGEO R GEO repository.Supplementary Bioinformatics online.