Table 2

EzArray has built-in algorithms that help users select statistical testing methods based on the number of sample groups and replicates

Sample Groups

Sample Replicates

Basic Statistical Method

Multiple Testing Procedure

Gene (feature) Limit Options

Main result files

1

Any

None

None

None

None

2

< 3 in either group

• Average values of each gene (feature) within group

• Calculate fold change of each gene between groups

None

1. Fold Change

2. Total number of genes

3. User gene list

• Gene list

• Fold changes

• Expression values

• Heatmap of top genes

≥3 in both groups

• two-sample Welch t-test (unequal variances)

• two-sample t-test (equal variances)

• standardized rank sum Wilcoxon test

• paired t-test

• Options for Raw/Nominal p-value calculation:

- Parametric

- Permutation

- Options for Side/Rejection Region: abs, upper, lower

• Bonferroni single-step FWER

• Holm step-down FWER

• Hochberg step-up FWER

• Sidak single-step FWER

• Sidak step-down FWER

• Benjamini & Yekutieli step-up FDR

• Benjamini & Hochberg step-up FDR – selected

• Storey q-value single-step pFDR

• Westfall & Young maxT permutation FWER

• Westfall & Young minP permutation FWER

• Fold Change

• Limit to

- Total number of genes

- adjusted p-values

- raw p-values

- test statistics

• User gene list

• Gene list

• Fold changes

• Statistic

• Raw p-values

• Adjusted p-values

• Expression values

≥3

≤3 in any group

• Calculate percentile of standard deviation (SD) of each gene cross all samples

• Select genes by a SD percentile cutoff

• Standard Deviation (SD)

• Total number of genes

• User gene list

• Gene names

• Standard Deviation

• Expression values

≥3 in all groups

• F-test

• Block F-test

• Options for Raw/Nominal p-value calculation:

- Parametric

- Permutation

- Options for Side/Rejection Region: abs, upper, lower

• Bonferroni single-step FWER

• Holm step-down FWER

• Hochberg step-up FWER

• Sidak single-step FWER

• Sidak step-down FWER

• Benjamini & Yekutieli step-up FDR – selected

• Benjamini & Hochberg step-up FDR

• Storey q-value single-step pFDR

• Westfall & Young maxT permutation FWER

• Westfall & Young minP permutation FWER

• Total number of genes

• User gene list

• Gene names

• Statistic value

• Expression values

While users can specify the statistical testing methods and parameters for the analysis, EzArray has built-in logics to select methods and parameters in the default modes. In general: 1) RMA is used to pre-process data; 2) If there is only one sample group, no analysis is performed; 3) If there are two groups and one or both of them has less than three replicates, then a fold change cut-off is applied, and the system returns the top genes (both up and down regulated, up to 100 in total) with 1.8 or higher fold change; 4) If there are three or more groups and one or more of them have less three replicates, then a Standard Deviation (SD) cut-off is applied, and the system returns the top genes (both up and down regulated, up to 100 in total) with SD within 90%; 5) If there are two groups and both of them have three or more replicates, then t test and Benjamini & Hochberg step-up FDR (BH) are applied, and the system returns the top 100 genes ordered by adjusted p-values, unadjusted p-values, and test statistics and with fold change >= 1.8; 6) If there are three or more groups and all groups have three or more replicates, then F-test and BH are applied, and the system returns the top 100 genes ordered by adjusted p-values, unadjusted p-values, and test statistics.

Zhu et al. BMC Bioinformatics 2008 9:46   doi:10.1186/1471-2105-9-46