countTransformers (version 0.0.6)

r2Transformer: Root Based Count Transformation Minimizing Sum of Sample-Specific Squared Difference

Description

Root based count transformation minimizing sum of sample-specific squared difference.

Usage

r2Transformer(mat, low = 1e-04, upp = 1000)

Arguments

mat

G x n data matrix, where G is the number of genes and n is the number of subjects

low

lower bound for the model parameter

upp

upper bound for the model parameter

Value

A list with 3 elements:

res.delta

An object returned by optimize function

eta

model parameter

mat2

transformed data matrix having the same dimension as mat

Details

Denote \(x_{gi}\) as the expression level of the \(g\)-th gene for the \(i\)-th subject. We perform the root and voom transformation $$y_{gi}=\frac{x_{gi}^{(1/\eta)}}{(1/\eta)}$$, The optimal value for the parameter \(\eta\) is to minimize the sum of the squared difference between the sample mean and the sample median across \(n\) subjects $$\sum_{i=1}^{n}\left(\bar{y}_i - \tilde{y}_i\right)^2$$, \(\bar{y}_i=\sum_{g=1}^{G}y_{gi}/G\) and \(\tilde{y}_i\) is the median of \(y_{1i}, \ldots, y_{Gi}\), and where \(G\) is the number of genes and \(n\) is the number of subjects.

References

Zhang Z, Yu D, Seo M, Hersh CP, Weiss ST, Qiu W. Novel Data Transformations for RNA-seq Differential Expression Analysis. (2019) 9:4820 https://rdcu.be/brDe5

Examples

# NOT RUN {
library(Biobase)

data(es)
print(es)

# expression set
ex = exprs(es)
print(dim(ex))
print(ex[1:3,1:2])

# mean-median before transformation
vec = c(ex)
m = mean(vec)
md = median(vec)
diff = m - md
cat("m=", m, ", md=", md, ", diff=", diff, "\n")

res = r2Transformer(mat = ex)

# estimated model parameter
print(res$eta)

# mean-median after transformation
vec2 = c(res$mat2)
m2 = mean(vec2)
md2 = median(vec2)
diff2 = m2 - md2
cat("m2=", m2, ", md2=", md2, ", diff2=", diff2, "\n")
# }