TY - GEN
T1 - Sensitivity, specificity and prioritization of gene set analysis when applying different ranking metrics
AU - Zyla, Joanna
AU - Marczyk, Michal
AU - Polanska, Joanna
N1 - Publisher Copyright:
© Springer International Publishing Switzerland 2016.
PY - 2016
Y1 - 2016
N2 - Microarrays were a trigger to develop new methods which can allow to estimate disturbances in signal cascades, characterized by sets of genes, in various biological conditions. Existing approaches of gene set analysis take information if genes are differentially expressed or are based on some gene ranking. The most commonly used method is Gene Set Enrichment Analysis (GSEA), where an assumption of uniform distribution of genes in some gene set is tested by weighted Kolmogorov-Smirnov test. Many studies present different gene set analysis methods and their comparison, however none of them focus on basic but crucial parameters, like the rank metric. In this paper we compare nine ranking metrics in terms of sensitivity, specificity and prioritization of identification of functional gene sets using a collection of 34 annotated microarray datasets. We show that absolute value of default GSEA measure is the best ranking metric, while the Baumgartner-Weiss-Schindler test statistic is the best statistical-based metrics, which can be used in Gene Set Enrichment Analysis.
AB - Microarrays were a trigger to develop new methods which can allow to estimate disturbances in signal cascades, characterized by sets of genes, in various biological conditions. Existing approaches of gene set analysis take information if genes are differentially expressed or are based on some gene ranking. The most commonly used method is Gene Set Enrichment Analysis (GSEA), where an assumption of uniform distribution of genes in some gene set is tested by weighted Kolmogorov-Smirnov test. Many studies present different gene set analysis methods and their comparison, however none of them focus on basic but crucial parameters, like the rank metric. In this paper we compare nine ranking metrics in terms of sensitivity, specificity and prioritization of identification of functional gene sets using a collection of 34 annotated microarray datasets. We show that absolute value of default GSEA measure is the best ranking metric, while the Baumgartner-Weiss-Schindler test statistic is the best statistical-based metrics, which can be used in Gene Set Enrichment Analysis.
KW - Functional enrichment efficiency
KW - Gene set analysis
KW - Ranking metrics
UR - https://www.scopus.com/pages/publications/84976354225
U2 - 10.1007/978-3-319-40126-3_7
DO - 10.1007/978-3-319-40126-3_7
M3 - Conference contribution
AN - SCOPUS:84976354225
SN - 9783319401256
T3 - Advances in Intelligent Systems and Computing
SP - 61
EP - 69
BT - 10th International Conference on Practical Applications of Computational Biology and Bioinformatics
A2 - Fdez-Riverola, Florentino
A2 - De Paz, Juan F.
A2 - Rocha, Miguel P.
A2 - Mayo, Francisco J. Domínguez
A2 - Mohamad, Mohd Saberi
PB - Springer Verlag
T2 - International Conference on Practical Applications of Computational Biology and Bioinformatics PACBB, 2016
Y2 - 1 June 2016 through 3 June 2016
ER -