.fi
.ll 60
.pl 5
.ce
PRE-REFINEMENT

    This process refers to what is done to the RNA 3D structure immediately
after it is obtained from the 3D template by appropriately substituting
double stranded helices corresponding to the stipulated basepair list.

    It will be recalled that generation of a 3D structure from the given
base pairing information is to first produce a structure that roughly
corresponds to what the 3D molecule would look like if its stems were
unwound, tertiary bonds broken, and the backbone (chosen to correspond to
the C1' atoms) laid out flat.  The next step is to substitute 3D double-
stranded helices into this template while maintaining the single-strand
connections
between the stems.  The result of this step is a 3D version in which all
the basepairs within a stem are correctly aligned with respect to one another
and the stems are interconnected with single strands which are only
approximately correct relative to base stacking and with respect to
the covalent bonding distance between successive bases. This approximation
also applies to the single strands comprising hairpin loops.  There is
therefore a need to improve the single strand portions of the initial
3D structure. This is done in the pre-refinement step which is
automatically invoked when the helix substitutions are completed.  The user
only sees the result.

    Initial schemes of pre-refinement were aimed at producing
"realistic" hairpin loops by building a library of these from previous
full scale refinements, that is, by doing energy minimization on select
ones and storing the results.  But this did not prove to be practical
because of the sensitivity of hairpin loop structure to composition.
Too large a library would be required.  We therefore treat them like
any other single strand during pre-refinement which consists only of
insuring that the distances between successive C1' atoms (the backbone
atoms) is that corresponding
to what it would be for the C1' atoms of a pair of stacked bases,
and is accomplished by a form of reduced dynamics.  Although
this technique does not insure that O3'-P bond distances are
correct within a single strand, it nevertheless seems to provide
a reasonable and safe starting point for user-selected further
refinements.  This is especially the case when compactification
is employed (invoked by default) which in effect reduces the length
of single strands by promoting single-strand base stacking.

.ce
THE END
