Removing Confounding Factor Effects using Linear Regression

Removing Confounding Factor Effects using Linear Regression

cfr_linReg(mat, cov)

Arguments

mat

a numeric matrix with the expression values, where columns are the samples and rows are probesets, transcripts, or genes.
cov

a list with covariates to regress out from the expression matrix

Value

a list with three elements, i) corrected expression matrix, ii) coefficients for the covariates and iii) p values for each covariate

Examples

myexp <- sapply( 1:10, function(i){ rnorm(n = 10000, mean = sample(1:3, 1), sd = sample(c(1, 3), 1)) })
mycov <- list(batch = sample(c(0, 1), 10, replace = TRUE))
resx <- cfr_linReg(myexp,mycov)
print(myexp[1:5,1:5])
#>          [,1]      [,2]      [,3]       [,4]       [,5]
#> [1,] 3.327853 -1.933807  3.214582 -0.2793156  1.3414952
#> [2,] 1.804408  3.096114  6.207055  3.2489682 -0.7878048
#> [3,] 1.989951  3.604186  3.220499  2.8413460  3.1144246
#> [4,] 2.822292  2.778184 -2.487390  1.5423143  2.6529450
#> [5,] 1.192382 -2.589336  7.016897  2.8087686  0.7772838
print(resx$correctedExp[1:5,1:5])
#>            [,1]       [,2]      [,3]       [,4]       [,5]
#> [1,]  3.1903319 -2.0713279  3.077061 -0.4168364 -0.1909669
#> [2,] -2.2191724 -0.9274665  2.183475 -0.7746124 -1.3373517
#> [3,] -0.3577359  1.2564986  0.872812  0.4936587  1.1698603
#> [4,]  1.6939186  1.6498107 -3.615764  0.4139409  0.9326460
#> [5,] -0.9940253 -4.7757435  4.830489  0.6223609 -0.1176258
dim(resx$correctedExp)
#> [1] 10000    10
head(resx$cov_coef)
#>      (Intercept)        cov
#> [1,]   1.5324621 -1.3949413
#> [2,]   0.5495469  3.4740337
#> [3,]   1.9445643  0.4031230
#> [4,]   1.7202990 -0.5919256
#> [5,]   0.8949096  1.2914980
#> [6,]   0.8641718  1.9133033
head(resx$cov_p)
#>      (Intercept)        cov
#> [1,]   0.2912556 0.41455274
#> [2,]   0.6728053 0.04908203
#> [3,]   0.0188698 0.62451397
#> [4,]   0.1057890 0.61397433
#> [5,]   0.5806840 0.50670608
#> [6,]   0.5783433 0.31454629