On 12:10 Fri 22 May , Francis E Reyes wrote:
> Maybe this is more of a statistics/normalization question, but say you
> have the same molecule crystallized in two different states. How would
> you put their refined b-factors (directly from the pdb) on the same
> scale and say compare the b-factors of residues in the binding pocket?
> I figure some normalization scheme is involved, but are there any
> specific examples ? (quantile normalization? z-score?)
>
> Are there any papers that discuss interpretation of b-factors (being a
> crystallographic property) to the relative dynamics/flexibility of the
> residue in the crystal (and subsequently in solution if the residue is
> not a crystal contact?). I'm particularly interested in the dangers of
> such an interpretation.
No doubt this isn't the best approach, but it's given me useful results
consistent with limited proteolysis, and I can actually understand and
easily do the math. For each residue in each structure, I calculate the
ratio of that residue's B to the average B for that structure. Then I
just plot the two structures per-residue on a graph vs the ratios. I can
look for gaps between lines to indicate differences, or just look at the
patterns of the lines to look for more relatively disordered or static
regions.
The goal is using the structure's average B to normalize for the
difference between two structures, and use the ratio instead of a
subtraction to try accounting for the difference in absolute B ranges
between structures. For example, a structure with <B>=30 might have
individual B's ranging from 10 to 45, and a structure with <B>=100 might
have B's ranging from 30-150. For these cases, this calculation would
give you 10/30 ~ 30/100, and 45/30 ~ 150/100, whereas a simple
subtraction would give you 10-30 != 30-100, and 45-30 != 150-100.
--
Thanks,
Donnie
Donnie Berkholz
P. Andrew Karplus lab
Oregon State University
|