Is ignorant rambling, was: DNA Structure: An Age of Refinement...

Clive Delmonte clived at
Mon Feb 15 07:27:12 EST 1999

Dear Kevin

Thankyou for your helpful posting on experimental errors and their
consequences on the outcomes of subsequent refinements.  This certainly
deals with one possibility, though I had not really thought that
experimental errors, which I guess are very rare, would bear directly on my

Perhaps I could focus my actual concern with a few specific examples ?  (I
could cite many others.)

1)      Heinemann et al. (EMBO J Vol 11 (1992) 1931 - 1939) reported on
CCGGCGCCGG as a B-DNA frgament in space group R3 with a cell side of 5.407
nm containing 9 duplexes.  Examination of the structure factor file
downloaded from Brookhaven (ref. 1CGC) suggests that, apart from R3,
reflections which are systematically absent also follow (h-k)=/=3n when
l=3n, and (h-k)=3n when l=/=3n.  These conditions for systematic absence
would allow a choice of cell side reduced by root 3.

It is not a question of experimental error, it seems to me, but a question
of considering all the possibilities.

This would include running a refinement algorithm using a side of 4.682 nm.
Since an algorithm, such as SHELX, is neutral, and including any necessary
re-indexing, would not a coherent, but different structure result from the
refinement ?

2)      Drew et al. (J Mol Biol Vol 125 (1978) 535 - 543) reported on
d(CGCG) x 2 (apparently a fragment of Z-DNA) crystallising in P6(1)22 with
a=b=3.08 nm (Brookhaven ref. 1ZNA).  They report on page 538:

"If helices A, B & C were translationally equivalent, then reflections with
h-k=/=3n would be extinguished, and a reduced cell with a=b=1.78 nm would
result. Although these reflections are not absent, they are systematically
weaker than those reflections for which h-k=3n. Every 3rd row in Fig.1(a) is
especially strong.  Hence helices A, B & C are not far from being

There is another possibility though.   That is, that helices A, B & C are
not far from being equivalent in a unit cell which should properly have its
side taken as 1.78 nm, and that systematically weak presences have been
created, where they would otherwise have been absent, precisely because the
helices A, B & C are not quite equivalent in the smaller lattice.

Again we see not an error in technique, but another (unconsidered)
possibility.   If the smaller unit cell were taken as the basis of a
refinement, would not a different but acceptable structure be the outcome ?

After scrutinising dozens of structure factor files downloaded from
Brookhaven, covering most oligonucleotide research groups (except Rich's),
it seems to me that there are many examples of patterns of systematically
absent and systematically weak reflections which bear more detailed
examination to see how far smaller unit cells can properly be chosen.

My primary concern remains, therefore, and it is on this that I need your
guidance.   If the structure factor files were accompanied by unit cells of
suitably reduced cross section, perhaps with re-indexing as necessary, would
not an unbiassed refinement program like SHELX turn out a new, different,
but valid structure ?


To:   Dr. Kevin Cowtan, Structural Biology Laboratory, University of York,

More information about the Xtal-log mailing list