postProcessLinRegMatrix {CNVScope}

Postprocess linear regression matrix.

Description

Takes a linear regression matrix and sets infinites to a finite value, and changes the sign to match the sign of the correlation for each value.

Usage

postProcessLinRegMatrix(
  input_matrix,
  LM_mat,
  cor_type = "pearson",
  inf_replacement_val = 300
)

Arguments

input_matrix

The input matrix, which consists of bins and samples (no LM or correlation has been done on the segmentation values)

LM_mat

The linear regression matrix, with rows and columns consisting of bins and the values being the negative log p-value between them.

cor_type

The correlation type ("pearson" (linear), "spearman" (rank), "kendall"(also rank-based)). Rank correlations capture nonlinear relationships as well as linear. Passed to stats::cor's method parameter.

inf_replacement_val

the value for which infinites are replaced, by default 300.

Value

The output matrix, or if using slurm, the slurm job object (which should be saved as an rds file and reloaded when creating the output matrix).

Examples

inputmat<-matrix(runif(15),nrow=3)
colnames(inputmat)<-c("chr2_1_1000","chr2_1001_2000","chr2_2001_3000","chr2_3001_4000",
"chr2_4001_5000")
rownames(inputmat)<-c("PAFPJK","PAKKAT","PUFFUM")
outputmat<-matrix(runif(15),nrow=3)
outputmat<-cor(inputmat)*matrix(runif(25,-30,500),nrow=5)
diag(outputmat)<-Inf
postProcessLinRegMatrix(input_matrix=t(inputmat),LM_mat=outputmat,cor_type="pearson",
inf_replacement_val=300)