The design has two conditions and samples are unpaired

For example, you observe expression values in 10 wild-type and 10 mutant organisms. For each gene, separately compute the average expression in each condition and subtract the two averages. Differences on additive scale are ratios on original scale. Thus, the difference in means relates to the mean fold change on original scale and is therefore called log ratio. Log ratio scores are the simplest method to rank genes. In R, load for example package twilight by calling library(twilight) and type:

score1 <- twilight.pval(X,y,method="fc",paired=FALSE)

where "fc" stands for fold change.

The difference in means can be misleading if the gene-wise variances differ much. Consider again the example in Figure 18.2. Taking variability into account leads to the classical t-test score. Type:

score2 <- twilight.pval(X,y,method="t",paired=FALSE)

0 0

Post a comment