This tool can be used to identify modulators of previously identified QTL effects as we have shown here: Hypothesis-free identification of modulators of genetic risk factors

The source code is availible here: https://github.com/molgenis/systemsgenetics/tree/master/eQTLInteractionAnalyser

The last build can be downloaded here

Input data

We require a folder with the following files:

• A file with genotypes called: Genotypes.binary
• A file with expression of the eQTLs: Expression.binary
• A file with covariate data: Covariates.binary

Optionally a file with known eQTL effect used to correct the covariate data and a list of samples to include in the analysis.

Below the plain text version of the needed files is described. The following command can be used to convert these to the .binary formats.

java -jar ../eQTLInteractionAnalyser-1.2-SNAPSHOT-jar-with-dependencies.jar \
	--convertMatrix \
    -i inputfile.txt \
    -o output.binary \

:exclamation: Important the rows (variants) in the genotype file must correspond to rows (genes) in the expression file. This means that both must have an equal number of rows. If a variant is affecting two genes, this variant should be included twice in the dosage file. The sample order must also be exactly the same

Genotype dosage data

Tab-separated matrix with variants in rows and samples in columns.

eQTL expression data

Tab-separated matrix with genes in rows and samples in columns.

Covariate expression data

Tab-separated matrix with proxy gene expression and other other potential covariates such as PCs to test in rows and samples in the same order as in the other files in columns.

QTL file

An eQTL result file as produced by our QTL mapping pipeline.

Columns
PValue
SNPName
SNPChr
SNPChrPos
ProbeName
ProbeChr
ProbeCenterChrPos
CisTrans
SNPType
AlleleAssessed
OverallZScore
DatasetsWhereSNPProbePairIsAvailableAndPassesQC
DatasetsZScores
DatasetsNrSamples
IncludedDatasetsMeanProbeExpression
IncludedDatasetsProbeExpressionVariance
HGNCName
IncludedDatasetsCorrelationCoefficient
Meta-Beta (SE)
Beta (SE)
FoldChange
FDR

Samples to include file

Per line a sample to include. No heading.

Example command

java -Xmx80g -jar ../eQTLInteractionAnalyser-1.2-SNAPSHOT-jar-with-dependencies.jar \
    -i /inputfolder/ \
    -o /outputfolder/ \
    -e eQTLs.txt \
    -c gender MEDIAN_3PRIME_BIAS MEDIAN_5PRIME_BIAS GC PCT_INTRONIC_BASES \
    -c2 Batch1 Batch2 \
	-is includedSamples.txt \
    -n 20 \
    -pc 20 \
    -nt 8

Options

Short	Long	Description
-dif	--chi2sumDiff	Find chi2sum differences for each covariate between 2 consequtive interaction runs
-nn	--noNormalization	Skip all normalization step. n must be 1
-perm	--permute	Run permutation
-nt	--threads	Number of threads
-ncn	--noCovNormalization	Skip covariate normalization step. n must be 1
-ec	--eqtlsCovariates	Path to the eQTL file to correct covariates
-c	--cov	covariates to correct for using an interaction term before running the interaction analysis
-cf	--covFile	File containing the covariates to correct for using an interaction term before running the interaction analysis. No header, each covariate on a separate line
-sw	--swap	File containing the SNPs to swap
-e	--eqtls	Path to the eQTL file to test for interactions
-ch	--cohorts	Covariates to correct for without interaction term before running the interaction analysis
-i	--input	Path to the folder containing expression and genotype data
-cm	--convertMatrix	Convert matrix
-is	--includedSamples	Included samples
-it	--interpret	Interpret the z-score matrices
-snps	--snpsToTest	SNPs to test
-n	--maxcov	Maximum number of covariates to regress out
-o	--output	Path to the output folder
-c2	--cov2	Covariates to correct for without interaction term before running the interaction analysis
-ct	--covTest	Covariates to to test in interaction analysis. Optional, all are tested if not used
-s	--start	Start round for chi2sumDiff option
-pc	--numpc	Number of PCs to correct for
-thr	--threshold	Z-score difference threshold for interpretation