Object Oriented Microarray Library: Replicate Ratios

The replicate ratio module provides the definition of the replicate.ratio class. See the bottom of the page for an example of how the class can be used.

Parent Class

The replicate.ratio class is derived from the two.groups class. See the description of that class for the meaning of the attributes

one, two, fit, good.score, avg, dif, aname, bname, stats

good.coding, as.data.frame

New Attributes

reps

A slide.replicate object.

QC

A vector containing the differences in log ratios at replicate pairs.

bad.score

A vector of scores that can be used to determine if the replicate ratios are more variable than one would expect by random chance.

R

Vector of log ratios at the first spot of a pair.

r

Vector of log ratios at the second spot of a pair.

name

A character string.

Methods

replicate.ratio(s, name, which, ...)

Constructor. The fist (required) argument is a complete slide object. The optional name argument is used as a label. The optional which argument is an extractor; it defaults to svol.extractor. Remaining arguments are passed along to the extractor.

plot(x, folddiff, goodflag, badch, ccl, name, ...)

The arguments to the plot method have the same meaning they had in the parent class. In addition to the plots already produced by the parent class, this method adds a normal proabbaility plot (or qq-plot) of the good.score t-statistics along with a Baggerly-Coombes plot picking out the genes that are significantly different.

as.data.frame

This adds the QC and bad.score vectors to the data frame produced by the parent two.groups class.

Description

The replicate.ratio class is the heart of the basic analysis of a single CG4 microarray. Given a complete.slide object, we first construct the replicates (reps) for the individual channels as an object of class slide.replicate. Next, we compute the replicate log ratios (R, r), the average log intensity (avg), the difference in log ratios (dif) and the average log ratio (ALR). Finally, the two smoothed estimates of standard deviation as a function of mean log intensity are merged (fit) and scores are computed for quality control (bad.score, which tells us if the replicate log ratios are more disparate than one would like) and for significant difference (good.score, which is a t-statistic that tells us if we believe that the gene represented by these replicate spots is truly differentially expressed in the two channels).