Need instructions for searching Tautomers.

Hi,  I need help finding a concrete example of how to do an indexed tautomer search using JCart. All I can find is scattered bits and pieces.  If you could read through this post, it may help assess how difficult the documentation is to follow, and as you read what I did, I hope you will see  how the lack of clear examples for specific cases really makes it hard to get anything done without ChemAxon support.

My understanding is that the most useful function for searching is jc_compare.  So I tried to learn to do a tautomer serch as follows:

So I took this to mean that "tautomerSearch:d" was how one would search for tautomers. 

Not really.

First of all specifying the default parameter for anything will not change anything - as a general rule in any API.

We perform a tautomer search by default (= automatically, no paramters set) only if the Tautomer Duplicate Filtering index option (or table option for JCB tables) was specified  during table/index creation. This method uses generic tautomers, and only applicable to duplicate filtering.

However in other cases it is also possible to perform a tautomer search, simply by specifying "tautomerSearch:y".

The relevant section of the JChem Database Concepts (a very useful document) details the various methods:

http://www.chemaxon.com/jchem/doc/dev/dbconcepts/index.html#tautomers

I was mentioning method #1 and #2 above, #1 is the most efficient method for duplicate filtering (t:d), #2 can be used for other searches (substructure, etc)

Best regards,

Szilard

Your suggestion:

However in other cases it is also possible to perform a tautomer search, simply by specifying "tautomerSearch:y".

does not work, It returns all whole bunch of stuff that is not a tautomer of the query structure.

I don't think you realize how unclear your documentation is.

I just need a concrete example of how to build an index to do a tautomer search, then how to search the index.

Below are some failure modes:

-------------------------------------------------------------------------------------

USING:

CREATE INDEX jc_idx ON structure(s_smiles) INDEXTYPE IS jchem.JC_IDXTYPE PARAMETERS('TABLESPACE=structure_indexes')

select s_smiles

  from structure

 where jc_compare(s_smiles,'C(Nc1ccnN1)c2cccs2','tautomerSearch:y') = 1;

returns

Brc1cc(sc1Br)C(=O)Nc2cc(n[nH]2)C3CC3

Brc1cc(sc1Br)C(=O)Nc2cc(n[nH]2)c3ccncc3

etc...

Which is obviously not right.

-----------------------------------------------

so then I tried the following>

does not work, It returns all whole bunch of stuff that is not a tautomer of the query structure.

In the example above you are performing substructure search (default) with the tautomer "y" option.

In this case tautomers of the query are enumerated, and substructure search is run with all tautomers (including the original).

Your query is a substructure of the targets, so the hits are correct.

CREATE INDEX jc_idx ON structure(s_smiles) INDEXTYPE IS jchem.JC_IDXTYPE PARAMETERS('TABLESPACE=structure_indexes tdf=y')


CREATE INDEX jc_idx ON structure(s_smiles) INDEXTYPE IS jchem.JC_IDXTYPE PARAMETERS('TABLESPACE=structure_indexes tdf:y')

Please see:

http://www.chemaxon.com/jchem/doc/dev/cartridge/index.html#create_index

PARAMETERS('param1=paramvalue1,param2=paramvalue2,...');

You need to use "," and "=" in the parameter list.

I hope this helps.

Best regards,

Szilard

Hi Szilard,

After our discussion today, I created an index without tdf=y:

CREATE INDEX jc_idx ON structure(s_smiles) INDEXTYPE IS jchem.JC_IDXTYPE PARAMETERS('TABLESPACE=structure_indexes,haltOnError=nf');

Once that was done, I tried :

select s_smiles
from structure
where jc_compare(s_smiles,'C(Nc1ccnN1)c2cccs2','t:d tautomerSearch:y') = 1;

and got the error:

ORA-29904: error in executing ODCIIndexClose() routine
ORA-20105: Tables created without tautomer duplicate filtering ("SENOBASE.JC_IDX_JCX") do not support duplicate search with --tautomer or --tdf option. Please change your table settings to using tautomer filtering(eg. using jcman GUI: File -> Table options -> 'Duplicate search uses tautomers') and recalculate your table for the most accurate tautomer duplication control.
ORA-06512: at "JCHEM.JCHEM_CORE_PKG", line 36
ORA-06512: at "JCHEM.JC_IDXTYPE_IM", line 668
ORA-29903: error in executing ODCIIndexFetch() routine
ORA-20105: Tables created without tautomer duplicate filtering ("SENOBASE.JC_IDX_JCX") do not support duplicate search with --tautomer or --tdf option. Please change your table settings to using tautomer filtering(eg. using jcman GUI: File -> Table options -> 'Duplicate search uses tautomers') and recalculate your table for the most accurate tautomer duplication control.
ORA-06512: at "JCHEM.JCHEM_CORE_PKG", line 36
ORA-06512: at "JCHEM.JC_I

..... But as we discussed before, If I create the index  with tdf=y, then it affects the behavior of all other searches.

How do I enable tautomer duplicate searches and regular searches on the same table, in a way that does not require me to specify tautomerSearch:n with every query that should not return tautomers??

Thank you,
~mike

Hi Mike,

I'm sorry, I totally forgot about this limitation ! 

1. You can go back to tdf:y index option - this also gives the best search performance as I have explained.

I would recommend this if you can put up with explicitely disabling tautomer search if not needed for your duplicate search - other searches are not affected.

2. Alternatively you may also use FULL search (t:f) instead of duplicate(t:d) (without the tdf index option)

Since you only have SMILES, which cannot contain query features, in your cases these modes are practicaly the same.

But as we discussed before, If I create the index  with tdf=y, then it affects the behavior of all other searches.

Only the duplicate (t:d) searches. 

Best regards,

Szilard

2. Alternatively you may also use FULL search (t:f) instead of duplicate(t:d) (without the tdf index option)


Since you only have SMILES, which cannot contain query features, in your cases these modes are practicaly the same.

One extra warning here: if you would like to simulate duplicate match with full structure search, you will also need to set other search options into "exact" mode.

For example: "t:f charge:e radical:e isotope:e stereoSearchType:e vagueBond:n"

See: http://www.chemaxon.com/jchem/doc/dev/cartridge/cartapi.html#jc_compare

Best regards,

Szabolcs

Thank you both.  Those explanations helped alot.  Everything is working as expected now.

Maya pointed out that t:ff also returns tautomers, but we'd sort of expect this behavior, because it's a duplicate match on a fragment.

~mike