6. Exporting Data
MAGNAMWAR statistical analysis can be exported into either tab-separated matrices or into graphical elements.
6a. Matrices
PrintOGSeqs
Allows the user to output all protein sequences in a specified OG in fasta format.
SurvAppendMatrix
When survival models are used in the MGWA a single .csv file is printed for each test. SurvAppendMatrix joins each individual file into one complete matrix.
WriteMCL
Writes the matrix result from AnalyzeOrthoMCL into a tab-separated file.
6b. Graphics
MAGNAMWAR also offers several options for graphical outputs. The several ways to visualize data are explained in detail below.
PDGPlot
PDGPlot visualizes the phenotypic effect of bearing a OG. The phenotype matrix is presented as a bar chart, and gene presence/absence is represented by different bar shading.
Calling PDGPlot requires the same pheno_data variable as AnalyzeOrthoMCL and similarly takes advantage of the user specified column names from the pheno_data data frame. It also requires the mcl_matrix object and specification of which OG to highlight.
For example, to visualize the means and standard deviations of the taxa which are present in OG “ex_r00002”, the OG with the lowest corrected p-value (“0.00186”). Green and gray bars represent taxa that do or do not contain a gene in the OG, respectively.
PDGPlot(pheno_data, mcl_matrix, OG = 'ex_r00002', species_colname = 'Treatment', data_colname = 'RespVar', ylab = "TAG Content")
The different taxa can be ordered alphabetically, as above, or by specifying the order with either the tree parameter (for phylogenetic sorting) or the order parameter, which takes a vector to determine order. For example, the taxa can be ordered by phylogeny by calling a phylogenetic tree file with the taxon names as leaves (note any taxa in pheno_data that are not in the tree file will be omitted).
tree <- system.file('extdata', 'muscle_tree2.dnd', package='MAGNAMWAR')
PDGPlot(pheno_data, mcl_matrix, 'ex_r00002', 'Treatment', 'RespVar', ylab = "TAG Content", tree = tree)
PhyDataError
PhyDataError is similar to PDGPlot and adds visualization of phylogenetic tree with the phenotypic means and standard deviation.
Calling PhyDataError requires the following parameters:
- a phylogenetic tree
- the same
pheno_datavariable asAnalyzeOrthoMCL(and its column names) - the
mcl_matrixobject - a specification of which OG to highlight
PhyDataError(tree, pheno_data, mcl_matrix, species_colname = 'Treatment', data_colname = 'RespVar', OG='ex_r00002', xlabel='TAG Content')
pdg_v_OG
PDGvOG produces a histogram of the number of OGs in each phylogenetic distribution group (PDG). A PDG is the set of OGs present and absent in the exact same set of bacterial taxa. The main purpose of the graph is to determine the fraction of OGs that are present in unique or shared sets of taxa, since the phenotypic effect of any OGs in the same PDG cannot be discriminated from each other.
To run PDGvOG, provide the output of FormatAfterOrtho. The num parameter (default 40) is used to specify the amount of OGs per PDG that should be included on the x axis.
For example, in the graph below there are 11 PDGs only contain one OG, meaning that 11 PDGs have one group that has a unique distribution of taxa present and absent.
PDGvOG(parsed_data,0)
Because this data is an extreme subset, it isn’t very informative. A full data set is shown below. This shows us that in this particular OrthoMCL clustering data, 4822 PDGs exist with only 1 unique OG in the PDG, and around 500 PDGs exist with 2 OGs that share the same presence and absence of taxa and so on.
QQPlotter
A simple quartile-quartile plot function that is generated using the matrix of AnalyzeOrthoMCL. To run QQPlotter, provide the output matrix of AnalyzeOrthoMcl.
QQPlotter(mcl_matrix)
Using this function, we provide a visualization of the use of the princ_coord parameter in AnalyzeOrthoMCL. Notice how each additional principal coordinate reduces statistical inflation. In practice it is usually computationally inexpensive to run several iterations of the analysis with a different number of principal coordinates to test how their incorporation improves the statistical fit.
ManhatGrp
ManhatGrp produces a manhattan plot for visual analysis of the output of AnalyzeOrthoMCL. In a traditional genome wide association study, a Manhattan plot is a visualization tool to identify significant p-values and potential linkage disequilibrium blocks. A traditional Manhattan plot sorts p-value by the SNPs along the 23 chromosomes. In our Manhattan plot, we sort by taxa instead of by chromosome number as shown below. The lines that emerge in our plot show the clustered proteins across taxa (which because they are in the same OG would have the same p-value).
To run ManhatGrp, provide the output of FormatAfterOrtho and the output of AnalyzeOrthoMCL as shown below.
ManhatGrp(parsed_data, mcl_matrix)## ordering by protein id within taxa...
## creating manhattan plot...
## finished.Again because this data is an extreme subset, it isn’t very informative. Another full data set example is shown below.
