VMD-L Mailing List

From: Peter Freddolino (petefred_at_umich.edu)
Date: Sun Mar 21 2021 - 10:36:48 CDT

Hi Ramon,
Have a look at what a pdb file actually contains and I think you'll
understand. There are only three decimal places in those coordinates, so it
is pretty clear that the very first coordinate in your file (originally
-12.171550750732422) must be stored in the pdb file as -12.172 (you can
look in a text editor to confirm). Then, due to the way floating point
numbers work, that won't be read into memory precisely as -12.172, but is
converted to the nearest representable value, roughly -12.17199993133545.
If you find this part surprising, have a look at how IEEE 754 floating
point numbers work. You can save yourself from all manner of headaches by
understanding this. You may find this website enlightening:
https://urldefense.com/v3/__https://www.h-schmidt.net/FloatConverter/IEEE754.html__;!!DZ3fjg!rC_hTECH-v91bXi2bpozk4Af-RYfwbIvIfDZBgFflYCrtHQAazrZNmK0roLQmPhfJA$
Best,
Peter

On Sun, Mar 21, 2021 at 3:09 AM Ramon Mendoza Uriarte <
ramendoza_at_uchicago.edu> wrote:

> Hello Ashar and Axel,
>
> Thank you to the both of you for your comments. I was not aware the
> coordinates would change depending on the machine or to precession. I would
> have thought the coordinates would be unchanging once the output was
> written.
>
> Thank you,
> Ramon
> ------------------------------
> *From:* Ashar Malik <asharjm_at_gmail.com>
> *Sent:* Saturday, March 20, 2021 10:20 PM
> *To:* Ramon Mendoza Uriarte <ramendoza_at_uchicago.edu>
> *Subject:* Re: vmd-l: Why do the coordinates of my atoms change when
> writing to a pdb file?
>
> Hi Ramon,
>
> This is related to precision.
> Was the coor file generated on the same computer? If you stick with the
> same machine this will not happen.
>
> Depending on what you are doing - this may or may not be ignorable.
>
> Regards,
> /A
>
> On Sun, Mar 21, 2021 at 12:31 PM Ramon Mendoza Uriarte <
> ramendoza_at_uchicago.edu> wrote:
>
> Hello VMD Community,
>
> I loaded a molecule with its respective .psf and .coor files, where the
> .coor file was read as a NAMD Binary Coordinates file. I then wanted to
> write out the pdb file; for example, by doing the following in the
> TkConsole:
>
> set all [atomselect top "all"]
> $all writepdb new.pdb
>
> However, I found it surprising that this yields slightly different
> coordinates and box dimensions. For instance:
>
> From .coor: (-12.171550750732422, 5.363558769226074, -15.094483375549316),
> (-11.626058578491211, 4.931212425231934, -15.867215156555176),
> (-12.769146919250488, 6.002536773681641,
> -15.656786918640137) <--- A small subset of all the coordinates
> cellBasisVector1 87.59627914428711 0 0
> cellBasisVector2 0 88.96563339233398 0
> cellBasisVector3 0 0 97.37928009033203
> cellOrigin 0.1676909178495407 0.361166387796402
> 0.04699717089533806
>
> From new.pdb: (-12.17199993133545, 5.363999843597412,
> -15.093999862670898), (-11.62600040435791, 4.931000232696533,
> -15.866999626159668),
> (-12.769000053405762, 6.002999782562256,
> -15.656999588012695) <--- A small subset of all the coordinates
> cellBasisVector1 87.59599685668945 0 0
> cellBasisVector2 0 88.96599960327148 0
> cellBasisVector3 0 0 97.37999725341797
> cellOrigin 0.16769009828567505
> 0.36116769909858704 0.04699849709868431
>
> So far, I have tried a Google and VMD archive search; however, I have not
> determined why this occurred. Additionally, I am using VMD for
> LINUXAMD64, version 1.9.4a51. Any comments on why this occurred would be
> appreciated.
>
> Thank you,
> Ramon
>
>