LFCs (log2-fold-changes) were computed for each gene in each condition relative to a relevant control condition (e.g. H37Rv grown on 7H9).
The LFCs were normalized by dividing by the standard deviation for each condition. (so roughly, |LFC|>2 are the outliers, i.e. tails of the distribution)
Note that essential genes often have LFCs around 0.
Data for some genes is not available if they have less than 3 TA sites, or the dynamic range of LFCs is less than 2.0 across all conditions. (This is the case for many essential genes.)
'leverage' on each plot is a statistical score that indicates how
much the regression is influenced by a few outlier points; the
traditional threshold is indicated on the plots (3p/n, which is about
0.07 for this dataset), but in practice, cases with leverage > ~0.2
seem to be the least reliable (most misleading) correlation
coefficients. (An example of high leverage is Rv2525c vs pssA, which
has a leverage of 0.374; it has a r^2 of 0.306, but this is primarily
due to 3 data points.)
If you have any questions or comments, send them to Tom Ioerger (ioerger@cs.tamu.edu).