To elaborate further on software packages available that do a reasonable job of fitting small molecules and outputting reasonable dictionaries, one could consider Afitt from OpenEye Scientific Software. They have academic as well as commercial licenses to their software.
From: CCP4 bulletin board [mailto:[log in to unmask]] On Behalf Of Ian Tickle
Sent: Monday, 9 January 2012 11:40 PM
To: [log in to unmask]
Subject: Re: [ccp4bb] unusual bond lengths in PRODRG cif file
On 9 January 2012 06:13, Shveta Bisht <[log in to unmask]> wrote:
> Dear all
> I have generated a refmac cif file for a molecule using PRODRG. I used
> JME editor to draw the molecule and ran PRODRG online with the
> options: Chirality-Yes, Charges-Reduced and EM-Yes. Please check the
> attachments for the molecule drawing and the expected bond lengths as
> listed in the PRODRG generated cif file. There are some unusual
> values. I have listed them below along with the likely explanations
> (just guesses).
> DRG CAR CAK single 1.390 0.025
> DRG CAQ CAB single 1.390 0.025
> DRG CAT CAI single 1.390 0.020
> All three are C-C single bonds where one of the carbons belong to
> aromatic ring. This might lead to a short bond, although 1.39 seems to
> be too short for this.
> DRG CAI NAL double 1.340 0.022
> DRG CAP NAL single 1.340 0.022
> Here N and both the carbons are sp2 hybridization, so there can be
> delocalization of electrons. Thus both bonds (single and double) are
> of similar length.
> DRG CAP CAA double 1.530 0.025
> As mentioned above, if pi (unhybridized) electrons of CAP are involved
> in CAP NAL bonding, then CAP CAA double bond essentially becomes a
> single bond with 1.53 bond length.
> I need advise on the way I have run prodrg and the explanations for the results.
> Is it common to observe such values? Or it is due to the alternating single and
> double bonds in this structure.
I agree completely with your assessment, there are certainly some
"unusual" values, but I think this goes beyond "mistakes". I have run
your structure through our own dictionary auto-generation software
which essentially takes suitable fragments from the CSD and stitches
them together (similar to CCDC-MOGUL but it doesn't use MOGUL and the
pieces are smaller).
Here are the values I get for comparison:
> DRG CAR CAK single 1.390 0.025 1.512(10)
> DRG CAQ CAB single 1.390 0.025 1.498(7)
> DRG CAT CAI single 1.390 0.020 1.449(24)
> DRG CAI NAL double 1.340 0.022 1.273(13)
> DRG CAP NAL single 1.340 0.022 1.350(28)
> DRG CAP CAA double 1.530 0.025 1.329(21)
so essentially backing up what you said.
My suspicion is that it's not generally recognised the central role
played by bond length & angle restraints in MX refinement (after all
if they weren't important then we wouldn't need to use them!), and as
a consequence some people seem to think that their exact values are
unimportant and "anything will do". Their importance soon becomes
obvious (particularly of course at medium-low resolution) if you try
to do unrestrained refinement: then you can easily get deviations 10
times greater than you get with restraints. So personally I believe
in using the most accurate restraints possible and "anything will NOT
My main concern would be that grossly inaccurate values such as these
find their way into the PDB and then people base their restraints on
these structures (though it's not a great idea to use the PDB
co-ordinates as a source of restraints!).
Before you ask, unfortunately our software is not available for
distribution but I can recommend GRADE from Global Phasing which gives
very high quality restraints; there's also of course JLigand from
CCP4, though I haven't used it personally and so cannot attest for the