Description

Set parameter useModel = modelLINEAR in the call of Matrix_eQTL_main to indicate that the effect of genotype on expression should be assumed to be additive linear.

References

The package website: http://www.bios.unc.edu/research/genomic_software/Matrix_eQTL/

See Also

See Matrix_eQTL_engine for reference and sample code.

Examples

library('MatrixEQTL')	

# Number of columns (samples)
n = 100;

# Number of covariates
nc = 10;

# Generate the standard deviation of the noise
noise.std = 0.1 + rnorm(n)^2;

# Generate the covariates
cvrt.mat = 2 + matrix(rnorm(n*nc), ncol = nc);

# Generate the vectors with genotype and expression variables
snps.mat = cvrt.mat %*% rnorm(nc) + rnorm(n);
gene.mat = cvrt.mat %*% rnorm(nc) + rnorm(n) * noise.std + 0.5 * snps.mat + 1;

# Create 3 SlicedData objects for the analysis
snps1 = SlicedData$new( matrix( snps.mat, nrow = 1 ) );
gene1 = SlicedData$new( matrix( gene.mat, nrow = 1 ) );
cvrt1 = SlicedData$new( t(cvrt.mat) );

# name of temporary output file
filename = tempfile();

# Call the main analysis function
me = Matrix_eQTL_main(
  snps = snps1, 
  gene = gene1, 
  cvrt = cvrt1, 
  output_file_name = filename, 
  pvOutputThreshold = 1, 
  useModel = modelLINEAR, 
  errorCovariance = diag(noise.std^2), 
  verbose = TRUE,
  pvalue.hist = FALSE );
# remove the output file
unlink( filename );

# Pull Matrix eQTL results - t-statistic and p-value
beta = me$all$eqtls$beta;
tstat = me$all$eqtls$statistic;
pvalue = me$all$eqtls$pvalue;
rez = c(beta = beta, tstat = tstat, pvalue = pvalue)
# And compare to those from the linear regression in R
{
  cat('\n\n Matrix eQTL: \n'); 
  print(rez);
  cat('\n R summary(lm()) output: \n');
  lmodel = lm( gene.mat ~ snps.mat + cvrt.mat, weights = 1/noise.std^2 );
  lmout = summary( lmodel )$coefficients[2, c(1,3,4)];
  print( lmout )
}

# Results from Matrix eQTL and 'lm' must agree
stopifnot(all.equal(lmout, rez, check.attributes=FALSE))

[Package MatrixEQTL version 2.1.1 Index]