How to propagate an ID tag in SDF via JKlustor

I am running JKlustor as follows:

jklustor P10635a.sdf -o wrclus:smi:fw.smi  -o "wrmols:sdf:cluster_*.sdf" -c sphex:0.85

The file P10635a.sdf has 71 molecules, each tagged with an <ID> tag  as is normal in an SDF file.

File is attached, but the problem seems to exist with any SD file I try.

The output clusters cluster_*.sdf do not contain this <ID> tag.  I would like to propagate this <ID> tag into the clustered output. 

Question: how do I do this please?

JKlustor identifies itself as v0.07.  (Maybe that should have been my first clue?) 

Thanks

John

Hi Guys - 

Can I re-phrase my question:  I want to have the identifiers of the molecules in the clustered results. Is there a way to do this please?  At the moment, there does not appear to be any way to trace back from the clustered results to the input molecules.

Thanks

John

Dear John,

 

Sorry for the late answer. Currently it is not possible to propagate input ids or other properties in jklustor. Implementing this functioanlity is in our plans, however it is not scheduled yet.

Using molconvert's canonic smiles functionality and simple bash tools a workaround can be constructed to assign input ID-s to generated cluster members. Overview:

• Create a file containing the input structures in canonic SMILES format and IDs separated by white space (chemaxon toolkits will interpret IDs as molecule name this way)


• For a selected cluster convert cluster members into canonic SMILES format and call grep for each member on the input file. This way cluster member's will be represented in SMILES format with IDs attached


• Optionally convert these files into SDF using molconvert. This way input IDs will get propagated as structure names.

Details

• molconvert smiles:q P10635a.sdf > tmp.smiles
  will convert input structures to canonical SMILES format


• cat P10635a.sdf | grep CHEMBL > ids.txt
  creates a text file with IDs.


• paste tmp.smiles ids.txt > input.smiles
  creates the file described above


• jklustor P10635a.sdf -o  "wrmols:sdf:cluster_*.sdf" -c sphex:0.5 -v
  creates individual files for each clusters (note that -v used to turn on verbose and stricter dissimilarity threshold used to create multiple clusters)


• for i in ` molconvert smiles:q cluster_5.sdf ` ; do grep "$i" input.smi ; done > cluster_5.smiles  
  locates IDs for a specific cluster
  Output from this command looks like:
  Fc1ccc(cc1F)-c1ccc(cc1)S(=O)(=O)Cc1ccc2CCNCCc2c1        CHEMBL1077991
  Clc1ccc(cc1)-c1ccc(cn1)S(=O)(=O)Cc1ccc2CCNCCc2c1        CHEMBL1078267
  Fc1ccc(cc1)-c1ccc(cn1)S(=O)(=O)Cc1ccc2CCNCCc2c1 CHEMBL1078268
  Clc1ccc(cc1)-c1ccc(cc1)S(=O)(=O)Cc1ccc2CCNCCc2c1        CHEMBL1078344
  Fc1ccc(cc1)-c1ccc(cc1)S(=O)(=O)Cc1ccc2CCNCCc2c1 CHEMBL1078345
  CC(C)Oc1ccc(cn1)S(=O)(=O)Cc1ccc2CCNCCc2c1       CHEMBL1078461
  O=S(=O)(Cc1ccc2CCNCCc2c1)c1ccc(nc1)N1CCCCC1     CHEMBL1078463
  Fc1ccc(Nc2ccc(cn2)S(=O)(=O)Cc2ccc3CCNCCc3c2)cc1 CHEMBL1078464
  Fc1ccc(Oc2ccc(cn2)S(=O)(=O)Cc2ccc3CCNCCc3c2)cc1 CHEMBL1078555
  


• for j in cluster_*.sdf ; do echo "Processing $j" ; for i in ` molconvert smiles:q $j ` ; do grep "$i" input.smi; done > `basename $j .sdf`.smiles ; done
  an outer loop can be used to process all clusters
  
  

If you have further questions please do not hesitate to ask them

 

regards,

Gabor

Hi Gabor

Thanks for your advice. I got your workaround to work!

I appreciate your support.

John

Dear  Gabor , do You fix the problem at the moment with ID in an output SDF file?

Or I need to attach ID's to a clustered file in the way you write above?

May be exist more quick way... I have to clusterize more than 30 SDF files...

Dear Igor,

 

Sorry, this is not solved yet.

The workaround above might be extended with an outer for loop iterating through all the sdf files needed clustering.

The step "cat P10635a.sdf | grep CHEMBL > ids.txt" relies on a common prefix of IDs of sdf files.

Alternatively a simple awk script can be used to extract an sdf property:

grepsdfid.sh:

#!/bin/bash

awk '
    BEGIN { 
        nextNAME=0
    } {
        if ( $0 == "> <STRUCTURENAME>" ) {
            nextNAME=1
        } else if ( nextNAME != 0 ) {
            nextNAME=0
            print $0
        } else if ( $0 == "$$$$" ) {
            nextNAME=0
        }
    }
    ' 

Usage:

cat input.sdf | grepsdfid.sh > ids.txt

Regards,

Gabor