If a cluster is startlingly obvious, then it'll be clear to the
eye. However, it would be nice to apply a rigorous test to detect
whether it is occurring: my particular interest is to show
that it is not occurring.
What I have tried doing is to take the two sequences aligned to
the 3D structure (e.g IL-2 from the HSSP database,
human and rat).
All pairs of amino acids within 15 angstroms are then taken
(calculated as the square root of the sum of squares of
X1-X2, Y1-Y2, and Z1-Z2, which I hope is correct).
These pairs are subdivided into 3 groups: those which are
both conserved in man and mouse, those which are both diverged
in man and mouse, and those where one residue is conserved
and the other is diverged.
These are separately visualised using a 2-D plot of the residue
numbers against eachother (like the two-D lot used to visualise
3-D structure). Any clusters should be seen, but it
would be nice to put this on a statistical footing allowing a
formal test of whether a significant cluster(s) is seen. Has that
been done?
(My simplistic statistical summary for IL-2 is that the mean and
standard deviation are similar for these three groups.
If replacements clustered strongly, it would be expected that
there would be differences. Inspecting the histogram of distances
for each group would also be informative, but the problem with
formal statistical testing is that the pairs of residues are not
independent).
Denis
(Ser).
Denis Shields, Genetics Department, Trinity College, Dublin 2, Ireland.