User d028dca803
25-06-2009 13:53:02
Hi,
I have a database of compounds in sdf format and I’m trying to generate a corresponding fingerprint database. Some compounds don’t get processes by generatemd and can’t be opened by JChem viewer – I’ve attached an example. This will open with other molecular viewing software. I can’t see any problem with the bond topology (is this the only data read to generate the fingerprint?).
Any suggestions greatly appreciated.
Gareth
ChemAxon e500b51457
25-06-2009 14:28:13
Hello,
Your SDF file is not correct. In the Bond Block in your file the third column is for the bond type.
Legal values are between 1 and 8. There is a 0 in the third column, which causes the error.
See the specification of the Bond Block in this document:
http://www.mdli.com/downloads/public/ctfile/ctfile.jsp
/>
Regards,
Erika
User d028dca803
26-06-2009 11:11:21
Thanks for the help.
So, generatemd requires bond type information. Is the minimal data then atom type connectivity and bond type? I ask this because the bond lengths are often wrong in sdf files.
ChemAxon e500b51457
29-06-2009 07:41:32
My colleague, who can answer this, is on holiday, he will answer to you soon.
Best regards,
Erika.
ChemAxon efa1591b5a
30-06-2009 14:10:36
Hi,
bond length is not needed, as our fingerprints (including the chemical fingerprint and the pharmacophore fingerprint) are topological (ie. connectivity based v.s. geometry based). But exact bond information, i.e. connectivity plus bond type are needed.
Some of our applications, like molconverter (and apparently our library of developers tool) have the capability to perceive bond and bond type. However, at present this feature is limited to those structure file types that cannot represent bond types (like PDB, for instance). Bear in mind though, that this bond/bond type recognition works for 3D structures.
HTH
Miklos