Phenotype Simulator

Phenotype Simulator

Our software package also includes the eQTL simulator we have used for the power simulations in our paper. In a nutshell, each gene is simulated as a linear function of all causal genes, the cis SNP, noise and hidden confounding factors. For more information, we refer to the supplement of our paper.

The simulator has a command line interface that is easily accessible:

./GNetLMM_simPheno --bfile bfile --pfile pfile

where

• bfile is the basename of of the binary bed file (bfile.bed,bfile.bim,bfile.fam are required).
• pfile is the basename of the phenotype file that is generated.The file pfile.matrix contains the gene expression values: each gene is saved in a row and each sample in a column. The file pfile.rows contains meta-information over the genes, pfile.cols meta-information over the samples. The file pfile.conf saves the hidden confounding factors, pfile.Aconf the mapping between the confounding factors and the genes, and pfile.Agene the gene-gene network, where A[i,j]=1 if gene j causes gene i.

By changing the following parameters different genetic architectures can be simulated and, in particular, the power experiments of our paper can be reproduced.

Option	Default	Datatype	Explanation
--seed	0	int	seed for random number generator
--networkDesign	sparse	string	can either be star or sparse
--T	100	int	number of simulated genes
--varSnp	0.1	float	variance explained by the causal cis marker
--varNetwork	0.8	int	variance explained by the gene-gene and gene-confounder network
--alpha	1.0	float	ratio between variance explained by the gene-gene network and the gene-confounder network
--nConfounder	3	int	number of confounders
--confPerGene	1	float	expected number of confounders per gene
--expN	5	float	expected number of in- and out-going edges