[Molvis-list] Meaning (or lack of it) in 5-place atomic coordinates

Gale Rhodes rhodes at usm.maine.edu
Tue Mar 8 13:18:09 EST 2005

Hi, Inbal!

Just for your information, although modeling programs might construct 
theoretical models and compute atomic coordinates with 5-point 
precision, no atomic coordinates from experimental structure 
determination (crystallography or NMR) are known to 5-point precision. 
Even three decimal places greatly overstates the certainty of atomic 
positions in experimental models.

For experimental macromolecular models at the highest resolution, if 
coordinates are expressed in angstroms, the first decimal place (tenths 
of angstroms) is the last significant digit, and thus it is uncertain 
(plus or minus 0.1 to 0.3 angstroms), while subsequent digits are 
meaningless. Uncertainty in atomic positions in the best experimental 
macromolecular models is usually not less than 0.1 angstrom. In most 
published models, it is higher.

In addition, uncertainty in atomic positions is not the same in all 
parts of an experimental model. And finally, these numbers treat the 
model as static, which is certainly not true. The coordinates are best 
viewed as centroids of motion, not as fixed positions. As a rule of 
thumb, I consider any intramolecular measurement on even the best 
macromolecular model to be uncertain by 0.1 to 0.2 angstroms.

In crystallographic models, uncertainty in atomic positions is not the 
same as stated model resolution, which has to do with the highest angle 
at which reliable data were collected. Tolerance or uncertainty in 
atomic positions can be estimated  from measures of data quality in 
various resolution ranges. Graphics pictures of macromolecular models 
convey a misleading sense of certainty about atomic locations.

Some (a small percentage, but too many) of models in PDB are of very 
low quality and reliability. Just because a model is in the PDB does 
not mean it is a high quality model. For an enlightening (and 
startling) introduction to judging the quality of macromolecular 
models, see Gerard Kleywegt's excellent structure validation tutorials 



On Mar 8, 2005, at 12:48 AM, Inbal Tuvi-Arad wrote:

> Hello all,
>  Does anyone know how can I convert xyz files that have 5 decimal point
> precision to pdb files without loosing the precision?  All my attempts 
> to
> use conversion programs or Website creates pdb files with only 3 
> decimal
> point precision.
> Sincerely,
> Inbal Tuvi-Arad
> __________________________________
> Dr. Inbal Tuvi-Arad
> Department of Natural Sciences
> The Open University of Israel
> 108 Ravutski St., POB 808,
> Raanana, 43107, Israel
> Tel: 972-9-778-1773
> Fax: 972-9-778-0661
> Email: inbaltu at openu.ac.il
> __________________________________
> _______________________________________________
> Molvis-list mailing list
> Molvis-list at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/molvis-list
Gale Rhodes, Professor of Chemistry
University of Southern Maine
PO Box 9300
Portland, Maine 04104-9300

More information about the Molvis-list mailing list