- Analysis title
- geneXplain GmbH
- com.genexplain.analyses (geneXplain analyses)
Estimate differential expression using the gene expression mixture model of EBarrays
- Input table - Table with normalized measurement values
- Input log-base - Logarithmic base of input data
- 1. Condition / group name - Name for first condition / group
- 1. Columns - Columns assigned to first condition / group
- 1. Is control - Check to not create pattern for this group
- 2. Condition / group name - Name for second condition / group
- 2. Columns - Columns assigned to second condition / group
- 2. Is control - Check to not create pattern for this group
- 3. Condition / group name - Name for third condition / group
- 3. Columns - Columns assigned to third condition / group
- 3. Is control - Check to not create pattern for this group
- 4. Condition / group name - Name for fourth condition / group
- 4. Columns - Columns assigned to fourth condition / group
- 4. Is control - Check to not create pattern for this group
- 5. Condition / group name - Name for fifth condition / group
- 5. Columns - Columns assigned to fifth condition / group
- 5. Is control - Check to not create pattern for this group
- FDR cutoff - FDR level for critical value calculation
- Model family - Distribution family used in mixture model
- Output folder - Folder to store output tables
Please not that the first two groups (named Treatment and Control by default) are not optional and unnamed groups are not considered. One group needs to be marked as Control group.
Description of Method and Output
EBarrays estimates differential expression between specified conditions / groups.
This tool provides for differential expression analysis using the EBarrays package. The platform tool can compare up to five conditions / groups. The groups consist of columns of a data table that contains normalized measurement values, e.g. from a normalized microarry experiment.
EBarrays sets up a mixture model matching the specified groups. Differential expression is identified when components for a pattern describe the distribution of measurement values well. Then probe / gene values in the corresponding group were significantly different from their values in the other groups. This is reflected by high posterior probabilities in the column named after that group.
The package estimates a critical posterior probability cut-off for the given FDR level on the basis of the fitted mixture model. Probes / genes exceeding this cut-off in some condition / group are indicated by a value of 1 (instead of -1) in the output column named "condition name Sig". Hence, to isolate the targets differentially expressed in a condition of interest, e.g. condition named "treatment", filter the table for all rows with a value of 1 in the column "treatment Sig". The direction of differential expression can be derived from the fold change column "condition name FC", which contains the log2-fold changes.
It is necessary to provide a unique name for each group. Also, at least two data columns are required per group and one group needs to be marked as control group.
Besides the main output table containing differential expression estimates for each probe / gene, EBarrays provides two diagnostic plots named EBarrays CCV and EBarrays Marginal fit. These plots enable a judgment about whether assumptions of the approach hold and how well the fitted model represents the data (please refer to the documentation of the EBarrays Bioconductor package for further details).
Kendziorski, C.M., Newton, M.A., Lan, H., Gould, M.N. (2003). On parametric empirical Bayes methods for comparing multiple groups using replicated gene expression profiles. Statistics in Medicine 22:3899-3914.