Impute from a Down-Shifted Normal Distribution — impute_left

Performs row-wise imputation by drawing random values from a normal distribution that is shifted to the left and narrower than the distribution of observed values.

This method assumes that missing values are primarily from proteins with low abundance (i.e., below the detection limit). The default shift and scale values are based on those used in the Perseus analysis platform.

Usage

impute_left_dist(object, shift = 1.8, scale = 0.3)

Arguments

object: A SummarizedExperiment object containing data with missing values.
shift: A numeric value specifying how many standard deviations to shift the mean of the distribution for imputed values. Default is 1.8.
scale: A numeric value to scale the standard deviation of the distribution for imputed values. Default is 0.3.

Value

A SummarizedExperiment object with missing values imputed from a simulated low-abundance distribution.

Details

This imputation method should be applied to log-transformed data, as the underlying assumption of a normal distribution is more appropriate in log space. An exception is made for rows with only one or zero observed values, where missing values are imputed with 0.

Examples

# Create data with NAs, typically representing log-transformed values
raw_data <- data.frame(
  Gene = c("GENEA", "GENEB"),
  SampleA = c(25.1, 28.5),
  SampleB = c(25.5, 28.9),
  SampleC = c(NA, NA)
)

pd_obj <- create_protdata(dat = raw_data)

# For reproducibility of the random imputation
set.seed(123)

imputed_obj <- impute_left_dist(pd_obj)
#> Error: unable to find an inherited method for function ‘impute_left_dist’ for signature ‘object = "ProtData"’
cat("Data after imputation:\n")
#> Data after imputation:
print(imputed_obj@data)
#> Error: object 'imputed_obj' not found