IUBio GIL .. BIOSCI/Bionet News .. Biosequences .. Software .. FTP

Distance geometry

Dave Schweisguth dcs at proton.chem.yale.edu
Thu Feb 22 13:37:04 EST 1996


Hi all,

I've been running distance geometry on data generated from a toy model (a
7-bp RNA duplex) just to see how things go before I hand it real NMR data
from a real molecule. My protocol is essentially dg_full_embed.inp from the
manual (with dgsa.inp for cleanup). It works fine, but some things puzzle me:

- Metrization doesn't appear to improve the results; it may even degrade
  them. The former is not surprising since I'm using unrealistically complete
  and accurate constraints, but the latter seems wrong. Without metrization
  (lines 106 and 107 commented out) the DG structure looks approximately
  correct and SA nearly always refines it correctly. With metrization the DG
  structure doesn't look like RNA at all -- it looks spherical -- and SA
  refines correctly it somewhat less often. (SA from random coordinates,
  often works, so successful SA is at best a statistical measure of a good DG
  result.) Is this to be expected? Metrizing numbers of atoms other than 4
  changes $non_met_gap in the expected manner but produces the same strange
  structures.

- I tried using "mmdg group" to define each base (with its exocyclic atoms)
  as a rigid group. (Not to solve the preceding, but just because I thought
  it might be helpful.) The resulting DG structure is virtually
  superimposable on the structure generated without groups defined (and with
  the same seed). My distance constraints don't include intrabase distances,
  so it's not that the distances which I think I'm defining with "mmdg
  groups" are redundant. All I did to my DG protocol to add groups was read a
  template structure in at the beginning and add a "@groups.inp" before the
  "storebounds". groups.inp parses without errors and the right number of
  atoms (about 12/base, depending on the base) are selected. Has anyone an
  example of "mmdg group" used correctly?

Any comments would be welcome. I used 3.1f.

Thanks,

--
| Dave Schweisguth                        For purposes of complying with    |
| dcs at proton.chem.yale.edu (MIME OK)      the New Jersey Right to Know Act: |
| http://proton.chem.yale.edu/~dcs/       Contents partially unknown.       |
| Yale Depts. of MB&B & Chemistry   Phone: 203-432-5208   Fax: 203-432-6144 |



More information about the X-plor mailing list

Send comments to us at archive@iubioarchive.bio.net