II) Dependence test: Myriads incorporates an autocorrelation test based on the generalized Durbin-Watson statistic. The test permits to identify strong dependencies in the p-values.
III) Simulator: Myriads includes the possibility of simulating lists of correlated p-values obtained by conducting
i) A two sample t-test for each gene. The size of the effects is user-defined by introducing the desired power for each test.
Detailed information on the different features is given in the program manual.
I) Multiple testing correction methods
As for version 1.2 Myriads performs the following methods: Bonferroni, Sequential Bonferroni (Holm 1979) and Hommel (Meijer et al 2019) FWER-based methods; Benjamini-Hochberg (1995), Benjamini & Yekutieli (2001) and Benjamini, Krieger & Yekutieli (Benjamini et al 2006) FDR-based methods, SGoF (Carvajal-Rodriguez et al. 2009), SLIM (Wang et al. 2011), Bon-EV (Li et al. 2017) and the permutation based step-down maxT procedure (Westfall and Young 1993). Myriads also estimates the proportion of true null hypothesis and computes the corresponding q-values (Tibshirani & Storey 2003; Storey et al 2004). The SGoF adjustment can also be computed without a list of p-values via interactive input of the number of comparisons and significant cases detected. The remaining methods need a list of probability values in a text file.
The input file is called by default PvalMyriads.dat but any other name can be given from the corresponding option in the menu or argument in the command line. The input file should have an integer number indicating the total number of tests and after it two columns with pairs of identifiers and p-values. The identifier can be a number or a character string.
See an Myriads input file example:
7 1 0.003 id2 5e-3 gen3 1.0 id4 0.34 5 0.01 6 0.004 7 0.445There is no limit on the number of p-values other than computer memory.
For performing the step-down maxT procedure a different input format is required. The normalized gene expression data format in Myriads consists in a text file, the first row has information about the class or response measurement (1 or 2) and the remaining rows have data, one row per gene. The first column contains the gene name/ID and the remaining columns represent the different samples.
Detailed information about the input formats, menu options and the command line arguments can be found in the Manual.
Run the program
a) Windows: double click the Myriads.exe file.
The program provides with different kinds of output files:
1) FWER_Summary_Myriads.txt and FDR_Summary_Myriads which includes only the tests remaining significant after the different adjustments.
The adjusted Bonferroni, BH, BK, BonSEV, BKY and SB p-values were computed following Yekutieli & Benjamini (1999). Adjusted Hommel p-values were computed following Meijer et al 2019. When the number of tests is below 10,000 (a myriad), the adjusted p-values for SGoF were computed following (Castro-Conde and de Uña-Álvarez, 2015). For higher number of tests, the SGoF metatest p-values are given by default, the adjusted p-values when computed, follow the approximation developed in (Carvajal-Rodriguez, 2018). Values under the SLIM column correspond to the q-value obtained using the SLIM estimate for the proportion of true nulls.
See an Summary_Myriads.html output file example: