Authors:
Irene Castro-Conde
and
Jacobo de Uña-Álvarez
Affiliation:
University of Vigo and Facultad de Económicas, Spain
Keyword(s):
Correlated Tests, False Discovery Rate, Gene Expression Levels, Monte Carlo Simulations, Multitesting Procedures.
Related
Ontology
Subjects/Areas/Topics:
Agents
;
Algorithms and Software Tools
;
Artificial Intelligence
;
Bioinformatics
;
Biomedical Engineering
;
Enterprise Information Systems
;
Genomics and Proteomics
;
Information Systems Analysis and Specification
;
Methodologies and Technologies
;
Operational Research
;
Simulation
Abstract:
In a recent paper (de Uña-Álvarez, 2012, Statistical Applications in Genetics and Molecular Biology Vol. 11, Iss. 3, Article 14) a correction of SGoF multitesting method for possibly dependent tests was introduced. This correction enhanced the field of applications of SGoF methodology, initially restricted to the independent setting, to make decisions on which genes are differently expressed in group comparison when the gene expression levels are correlated. In this work we investigate through an intensive Monte Carlo simulation study the performance of that correction, called BB-SGoF (from Beta-Binomial), in practical settings. In the simulations, gene expression levels are correlated inside a number of blocks, while the blocks are independent. Different number of blocks, within-block correlation values, proportion of true effects, and effect levels are considered. The allocation of the true effects is taken to be random. False discovery rate, power, and conservativeness of the meth
od with respect to the number of existing effects with p-values below the given significance threshold are computed along the Monte Carlo trials. Comparison to the classical Benjamini-Hochberg adjustment is provided. Conclusions from the simulation study and practical recommendations are reported.
(More)