Chemical filter terms and similarity search

Hi,

We have been testing the chemTermsFilter method for structure search options with some of the built in functions such as hasRadical, atomCount etc. So far this seems to be performing well but we have run across something troubling when used in conjuction with a similarity search.

With no filters applied we get, say, 500 hits for a search. When we apply a filter such as "atomCount() > 100" we get a smaller result, say 100, and when we run "atomCount <= 100" we get say 200 hits. Shouldn't these two result sets add up to the initial result count of 500? This only seems to happen (consistently) when doing a similarity search, filtered substructure results always add up to the unfiltered total.

We can provide some example result sets if that would help. We are running JChem 5.4.

regards

Richard

We are actually running JChem 5.7.0.

Richard

I've included a few attachments to demonstrate this issue. There are three files:

1. chem_smiles_unfiltered.out


2. chem_smiles_gt_100atoms.out


3. chem_smiles_lteq_100atoms.out

These correspond to the results we're receiving from Similarity searching with the following chemical terms filters:

1. None


2. atomCount()>100


3. atomCount()<=100

I've copied the expressions verbatim, so these should correspond to exactly what we're doing on the search.

For the three searches, we received counts of: 59 (unfiltered), 2 (>100), and 45 (<=100).

It's also worth noting that we've observed something similar with the hasRadical expression, where we get counts of 61 (unfiltered), 41(!hasRadical), and 3(hasRadical).

EDIT: changed "database" to "search".

hello,

we've reproduced the problem.

we will get back to you in the next few days when we will know more about the issue

The bug will be fixed in 5.11.2 due in the next 2 weeks.

Regards,

Robert

Hi Robert,

Any chance of an update regarding the expected delivery date for the new release of 5.7?

Thanks,

Jim

Hi Jim,

We still need few days. Many pieces of related new code was merged back in 5.7 but tests still fail.

Regards,

Miklos

Thank you for the update. We have currently rolled out a limited version of the filtering functionality with no support for similarity searching. We can roll out an update as soon as a new 5.7 jar is available, and of course would prefer a well tested version to avoid hitting any regressions.

I am writing to inform you that we have release JChem 5.7.3. 

https://www.chemaxon.com/download.php?d=/data/download/jchem/5.7.3/

This version solves the problems of similarity searches combined with chemical terms filtering expressions.

Kind regards,

Krisztina

Great, thanks for the update. We'll be deploying imminently.

Jim