Microarray diagnostics

Summary statistics

Raw data should be on the 2¹⁶ scale, with data ranges of (0, 65,535). Statistics of interest include, min, max,range, summary,# NA's, # saturated for each slide, or for each block with slides. For each experiment, the number of Empty, spots or positive/negative controls may be of interest from the annotation information.

Statistical measures

Diagnostic measues such as five number summary, Wikipedia:Quartiles, including measures such as Wikipedia:Skewness, and Wikipedia:Kurtosis.

Examples using apply

 # Ranges
 apply(RG$R, 2, range, na.rm=TRUE)
 apply(RG$G, 2, range, na.rm=TRUE)
 # Maximums
 apply(RG$R, 2, max, na.rm=TRUE)
 apply(RG$Rb, 2, max, na.rm=TRUE)
 apply(RG$G, 2, max, na.rm=TRUE)
 apply(RG$Gb, 2, max, na.rm=TRUE)
 # Examining backgrounds that are higher than foreground
 apply(RG$R < RG$Rb, 2, sum, na.rm=TRUE)
 apply(RG$G < RG$Gb, 2, sum, na.rm=TRUE)

There are several times where the data ranges, or the number of introduced missing values (NA's) can be investigated during background correction and normalization.

Background Correction

 #Number of missing values
 apply( backgroundCorrect(RG, method="subtract"), 2, sum(is.na))

Comparing channels

Differences between the Red and Green channels can be examined by plotting the differences in summary statistics, for example the pseudocode below plots the counts for the Green channel versus the Red channel where the backgroun is higher than the forground.

 plot(apply(RG$R < RG$Rb, 2, sum, na.rm=TRUE), apply(RG$G < RG$Gb, 2, sum, na.rm=TRUE), type="n")
 text(apply(RG$R < RG$Rb, 2, sum, na.rm=TRUE), apply(RG$G < RG$Gb, 2, sum, na.rm=TRUE), seq(colnames(RG)))