WISARD provides various types of meta-analysis, including simple to sophiscate one.
When your test result only provides p-value, a simplest meta analysis, Fisher's method can be applied. If your result also includes standard error (SE) or other weight information, a Stouffer's method can be additionally applied. For the FARVAT analysis, meta-FARVAT can be applied.
In order to perform a meta-analysis except for meta-FARVAT, the following requirements should be met.
Multiple results with same type of test are required. Each test should be written as row-wise (in other words, one test result per row).
The first line of each result should be its header and separator of each column should be (multiple) whitespace(s). 'Whitespace' is including only tab and space.
For each test, a column that uniquely distinguishes every row is required, in order to matching the result from same marker across multiple studies.
In default, only the markers that its results are fully available across all given results will be tested. In order to adjust this (for example, to perform a meta-analysis unless the result is available at least 2 files), --avail can be used.
WISARD will automatically seek appropriate columns according to the predefined rules. The rules are below.
For the p-values, a column named 'P' without quote or any column(s) that start with 'P_' without quote (such as P_REGR or P_QLS) will be included to meta-analysis.
For the weight which is required to perform Stouffer's method, a column named 'SE' or 'MAF' without quote will be searched. If the result file contains both column, an error will be raised by ambiguity. In this case, the weight column should be ignored via --metaweight without parameter or should be specified with --metaweight [column name].
For the marker name, a column named 'VARIANT' or 'SNV' or 'SNP' or 'MARKER' without quote will be searched. If the result contains at least two of the above column names, an error will be raised due to ambiguity. In this case, the marker name column should be specified with --metamarker [column name].
Unlike the other types of analyses in WISARD,
an input to perform meta-analysis is a set of results itself.
In other words, it is not possible to assign other type of input while the meta-analysis is performed. The following command is a simple meta-analysis example. In the below example, it is assumed that the both dataset sample1 and sample2 have continuous phenotype.
As shown in the third example, there is not input like the first and second command.
--meta option substitutes an input.
As a result, .meta.res will be produced with the following information.
Output index for extension [meta.res] is currently not available
If the marker column name is not one of the above,
the marker name column should be specified with --metamarker [column name].
In same manner, if the weight column name is not one of the above but if Stouffer's method is desired, the weight column name should be specified with --metaweight [column name].
FARVAT proposed by Choi et al. is an analysis method to perform gene-level test. Using the scheme of WISARD, it is possible to conduct meta-anaysis using multiple FARVAT results.
Meta-analysis using multiple FARVAT results can be done with the following procedure.
For the datasets of interested, perform FARVAT analysis with --makefarvat.
--makefarvat option lets WISARD to generate additional information to perform Meta-FARVAT.
Note that FARVAT produces .gene.res, those results should be given to perform meta-analysis.
Perform meta-FARVAT analysis, with --meta option. However, unlike the above meta-analyses, --farvat should be added. It indicates WISARD to perform meta-analysis as meta-FARVAT way.