The eqmr-fcr program carries out Bayesian regressions between cis-SNPs and feature expression levels (depending on which level was chosen with eqmr-fdb). Several genetic models can be used: additive, additive+dominance, recessive/dominance. The program also always performs standard linear regressions (additive model only).
Results of the regressions are stored into a binary database with one file per chromosome. The program can also be used to carry out permutations.
|--help||display a brief help on the program usage|
|--verbose||send messages on stdout to see what the program is doing|
|--qnorm||quantile normalize the expression levels|
|--pval||do permutations for p-values (otherwise Bayes factors)|
|--best||for permutations, report only the best SNP per gene (min p-val or max Bf)|
|-d||--data||the database (.db file)|
|-f||--feature||the feature database (.fdb file)|
|-o||--output||the output result database|
|-w||--window||the window size of the cis-region, in bp (default 500000)|
|-m||--model||the model for computing the Bayes Factors (configuration file)|
|-p||--permut||the number of permutations to perform (default is none)|
|-t||--feature-table||A table of features to look at|
|-c||--chrom||the name(s) of the chromosome(s) to focus on (default is all)|
The Bayesian regression approach allows to investigate several kinds of genetic models at the same time:
- purely additive;
- additive with moderate dominance:
- full dominance.
The models we want to compare and the prior distributions on the effect sizes need to be described in a configuration file. It is possible to assume that the effect sizes are drawn from a mixture of normal distributions centered on 0, as long as we precise their respective standard deviations.
As shown below in the examples, each genetic model corresponds to a section of the configuration file enclosed by tags: <a>...</a> for the purely additive model, <d>...</d> for the additive with moderate dominance model, and <r>...</r> for the recessive/dominance model. The first value after the tag (e.g. <a> 1) indicates the weight of the model. Then, within each section, the first column is for the weight on the corresponding prior distribution, and the second column gives its standard deviation. For the additive with moderate dominance model (tag <d>), the second column indicates the standard deviations for the additive effect, and the third column the standard deviations for the dominance effect.
We can access a short help with the following command:
$ eqmr-fcr --help eqmr-fcr - version 2.1 Copyright (C) 2008,2009 Jean-Baptiste Veyrieras (University of Chicago) eqmr-fcr comes with ABSOLUTELY NO WARRANTY. This is free software, and you are welcome to redistribute it under certain conditions. --help Display a brief help on program usage --verbose Output message on standard output to see what the program is doing --qnorm Quantile normalize the expression levels --pval For permutations, do permutation for p-values (otherwise bf) --best-snp For permutations, report only the best SNP per gene (min p-val or max bf) --data or -d The database file (.db file) --feature or -d The feature database file (.fdb file) --output or -o The stem name of the result database --window or -w The window size (bp) from either side of the gene (default is 500000) --model or -m The model for computing the Bayes Factors --permut or -p The number of permutations to do (default is none) --feature-table or -t A table of features to look at --chrom or -c The chromosome to focus on
Before launching the analysis, we need to write a configuration file, such as this one (here, we want to compare only the purely additive model and the additive with moderate dominance model, each with the same weight):
$ cat bayesreg.conf <a> 1 1 0.05 1 0.1 1 0.2 1 0.4 1 0.8 1 1.6 </a> <d> 1 4 1 0.05 0.0125 1 0.1 0.025 1 0.2 0.05 1 0.4 0.1 1 0.8 0.2 1 1.6 0.4 </d>
Then, the program can be launched, eg.:
$ eqmr-fcr -d data.db -f genes.fdb -o results_cisreg -m bayesreg.conf --qnorm --c chr22
It is also possible to compute the Bayes factor only for a subset of the features, listed in a file like this:
$ cat feature_table.txt chromosome feature chr22 ENSG00000100425 $ eqmr-fcr -d data.db -f genes.fdb -o results_cisreg -m bayesreg.conf --qnorm --c chr22 -t feature_table.txt