Postprocessing

Postprocessing

Concatenating Files

If the association updates have been computed blockwise, we have to merge the partial results to one file:

./GNetLMM/bin/gNetLMM_postprocess --concatenate --infiles infiles --outfile outfile

• infiles is a unique prefix describing the results files infiles.startTrait_*.csv from the previous step.
• outfile.csv is the filename of the merged results.

Merging Association Results

Finally, we merge the results from the initial association scan with the updated assocations:

./GNetLMM/bin/gNetLMM_postprocess --merge_assoc --assoc0file $ASSOC0FILE --assocfile $ASSOCFILE

• assoc0file is the basename of the initial association scan.
• assocfile is the basename of the updated association scan. The updates p-values are assumed to be saved in assocfile.csv for the method to be working. P-values are saved in assocfile.pv.matrix, SNP weights in assocfile.beta.matrix.

Plotting Power

Finally, we plot the power of GNetLMM versus a vanilla LMM:

./GNetLMM/bin/gNetLMM_postprocess --plot_power --assocfile assocfile --assoc0file assoc0file  --plotfile plotfile --pfile pfile --bfile bfile --window window

• assocfile is the basename of the GNet-LMM resultsfile.
• assoc0file is the basename of the LMM resultsfile.
• plotfile is the filename of the power plot.
• bfile is the basename of the bed file.
• pfile is the basename of the phenotype matrix. This method does only work for simulated phenotypes as it plots the false positive rate vs. the true positive rate. The true underlying gene-gene network must be deposited in pfile.Agene.
• window determines the minimal distance between the SNP and the gene to call the association a trans association.

Creating vstructure file in nice output format

The vstructure file can be converted to a more amenable format as follows:

./../GNetLMM/bin/gNetLMM_postprocess --nice_output --bfile bfile --pfile pfile --vfile vfile --outfile outfile --assoc0file assoc0file --assocfile assocfile --blockfile blockfile

• bfile is the basename of the bed file.
• pfile is the basename of the phenotype matrix.
• vfile is the basename of the vstructure results file.
• assocfile is the basename of the GNet-LMM resultsfile.
• assoc0file is the basename of the LMM resultsfile.
• blockfile is the basenome of the GNet-LMM resultsfile when the causal chain is blocked (conditioning on the anchor gene). This argument is optional.
• outfile is the basename of the nicely formatted vstructure results file. Each updated association corresponds to one row. The first column contains the anchor snp, the second column the anchor gene, the third column the focal gene, the fourth column the orthogonal gene(s), the fifth column the p-value without conditioning (vanilla LMM), the sixth column with conditioning on the orthogonal genes (GNet-LMM), the seventh column with conditioning on the orthogonal genes and the anchor gene (Block-LMM).