Project

General

Profile

Bug #211

pdb2gmx functionality broken - bond lengths changing

Added by Jussi Lehtola almost 12 years ago. Updated almost 12 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Erik Lindahl
Category:
mdrun
Target version:
Affected version - extra info:
Affected version:
Difficulty:
uncategorized
Close

Description

Created an attachment (id=284)
Methanol & undecanol structures with a ffoplsaa database

New pdb2gmx streches bond lengths (i.e. blows out atom coordinates) compared to Gromacs 3.3.3 . This can be seen by comparing e.g. the methanol molecule output of 3.3.3 vs that of 4.0_rc1.

Also, Gromacs 3.3.3 version is buggy for long alcohols; when visualizing with e.g. VMD one sees some false artefacts around the oxygen. You could maybe check this too?

attachments.tar.bz2 (7.68 KB) attachments.tar.bz2 Methanol & undecanol structures with a ffoplsaa database Jussi Lehtola, 09/25/2008 10:57 AM

History

#1 Updated by David van der Spoel almost 12 years ago

This is rather vague. Can you please supply an example? Definitely pdb2gmx is not made for these small molecules.

#2 Updated by Jussi Lehtola almost 12 years ago

They're included in the attachment.

Running

pdb2gmx -f metanoli.pdb -o metanoli.gro -p metanoli.top -ff oplsaa

you get the .gro file which is generated incorrectly: as compared to 3.3.3 e.g. the C-O distance is increased about twofold.

#3 Updated by David van der Spoel almost 12 years ago

I now see that you edited the rtp file.

If I do:

echo 0 0 | g_confrms -f1 metanoli.pdb -f2 metanoli-new.gro
I get
Root mean square deviation after lsq fit = 0.0480464

However, if I first run
pdb2gmx -f metanoli.pdb -o metanoli.gro -p metanoli.top -ff oplsaa
and then

g_confrms -f1 metanoli.pdb -f2 metanoli.gro

I get

Root mean square deviation after lsq fit = 0.000233141

In other words, it is the same. So I wonder how you produced the metanoli-new.gro file. I tried both pdb2gmx from 3.3 and 4.0.

#4 Updated by Jussi Lehtola almost 12 years ago

You are right, the bug is not in pdb2gmx.

What I do is:

1)
pdb2gmx -f metanoli.pdb -o metanoli.gro -p metanoli.top -i metanoli-posre.itp -ff oplsaa

Gromacs 3.3.3 : Root mean square deviation after lsq fit = 0.000345709
Gromacs 4.0_rc2 : Root mean square deviation after lsq fit = 0.000233141

2)
genbox -cp metanoli.gro -box 0.4 0.4 0.4 -o metanoli.gro

3.3.3 : Root mean square deviation after lsq fit = 0.000345709
4.0_rc2: Root mean square deviation after lsq fit = 0.000233141

3)
editconf -density 880 -f metanoli.gro -o metanoli.gro

3.3.3 : Root mean square deviation after lsq fit = 0.000824607
4.0_rc2: Root mean square deviation after lsq fit = 0.0480464

So it is editconf which is broken.

#5 Updated by David van der Spoel almost 12 years ago

Maybe the explanation in editconf manual is not clear enough. The density command is made to scale the box (and the coordinates) until mass/volume gives the right density. In the current example the mass is incorrectly determined from the gro file as well (Calpha is interpreted as a Calcium). I will update the help text in editconf.

#6 Updated by Jussi Lehtola almost 12 years ago

OK, but why on Earth does it scale the coordinates as well? Could you add an option not to scale the coordinates, just scale the box? (OK, this causes problems if the box is smaller than the molecule, but the current behavior causes some, too!)

#7 Updated by David van der Spoel almost 12 years ago

Well obviously if you scale the box down you get clashes with periodic images, and if you scale the box up you may break bonds. If you want to obtain a liquid with proper density the best way is to use genconf to make multiple copies of your box and then run MD with pressure coupling. If you just want to set the box vectors you can use the -box option.

Also available in: Atom PDF