program to combine omit maps?

[Gerard Kleijwegt]

In article <32gld0$d1q at mserv1.dl.ac.uk>, <HEFFRON at UCRAC1.UCR.EDU> writes:
|> 
|> Does anyone know of a program or procedure to combine several omit maps, each
|> omitting a different region, into one big map?  (We usually calculate maps with
|> X-PLOR, then make dsn6 files for O with a version of Mappage.)
|> 
|>                                  Thanks,
|>                                  Susan Heffron
|> 
|> --------------------------------------------------------------------------
|> Biochemistry Dept.                       phone:    (909) 787-4196
|> University of California, Riverside      internet: Heffron at ucrac1.ucr.edu
|> Riverside, CA 92521                      bitnet:   Heffron at ucrvms
|> U.S.A.
|> --------------------------------------------------------------------------
|> 

the simplest way is to calculate omit maps in xplor *not*
just around the omitted region, but around the entire
molecule (since you've gone through all the trouble
of a simul ann run already, you get the rest of the
map for free)

do this by putting the following at the end of your
sa-omit input file:

 { calc map }
 xrefine
  mbins 15
  update print r-factor
  do scale (fcalc=fobs)
  do amplitude (fcalc = 2*fobs - fcalc)
  map
   cushion=8.0 extend=molecule selection=(all)
   output=/nfs/scr_uu5/gerard/sa_omit_map.xplor {<==}
  end
 end

now you can simply combine the maps with MAPMAN (part
of the RAVE averaging package; on the O ftp server)

> read map1 filename x-plor
> read map2 filename x-plor
... (etc)

> add map1 map2
> add map1 map3
.. (etc)

note that the maps need not necessarily cover
the same part of space; only those bits of
map2 which also lie inside map1 will be added

at the end, map1 will be the sum of all omit maps;
you can then save it as a DSN6 file:

> mappage map1 dsn6_filename

or, since xplor maps tend to have sigmas of ~1 which
may give problems with mappaging and/or contouring,
you may first multiply the summed map by a constant:

> mult map1 100.0

probably, if you want to do things properly, you will
have to scale the maps together (since they were
calculated with different nrs of atoms); i guess
you would do this by dividing each of them by their
own variance prior to summing them

even this is simple in MAPMAN:

> list map1
(this will show the variance of the map, a.o.)
> divide map1 var_1

etc. for all the maps, then add, mult & mappage

back to your original question: combining maps
of small tidbits of your protein is a tad less
trivial because you will have to account for
the fact that these maps may have overlap which
(if you were to sum them without thinking)
would lead to junk density in these overlap
regions

the problem *can* be overcome, but it's an
awful lot of work (for the die-hards: i would
first convert each map into a mask; then use
MAMA (a mask-editing program, also in RAVE)
to get a set of non-intersecting masks; then
"average" each of the maps onto their trimmed
mask and then combine them; even then there
is a problem in determining the relative scales
of the maps)

in other words: calculate maps around your entire
molecule after the sa-omit & combine them with
MAPMAN

hope this helps

--gerard