Similarity specific search options

In this section search options concerning similarity are summarized and their usage is shown in different search interfaces.

Dissimilarity metrics

Sets the metric to be used during similarity search. MolSearch API


Not applicable.

JChemSearch API


        JChemSearchOptions searchOptions = new JChemSearchOptions(SearchConstants.SIMILARITY);
        searchOptions.setDissimilarityMetrics( "tanimoto" / "tversky" / "dice" /
                "euclidean" / "normalized_euclidean" / "substructure" /  "superstructure" );
        // ...
        JChemSearch searcher = new JChemSearch();
        searcher.setSearchOptions(searchOptions);

Default value is null, which means using the default metric.

JChem Oracle Cartridge


Use the jc_compare operator with dissimilarityMetric.

The values accepted as <metric> depend on the structureType.

    In case of table types anyStructures and molecules <metric> may be one of the following:
        tanimoto
        tversky
        euclidean
        ask for others
    In case of table type reactions <metric> may be one of the following:
        ReactantTanimoto
        ProductTanimoto
        CoarseReactionTanimoto
        MediumReactionTanimoto (default)
        FineReactionTanimoto

Example:

select count(*) from nci_1k where jc_compare(structure, 'C[C@H](cs)C(=O)N1CCC[C@H]1C(O)=O |r|',
        't:i dissimilarityThreshold:0.61 dissimilarityMetric:tversky;0.3;0.7') = 1;

jcsearch command line tool


Not applicable.

Dissimilarity threshold for similarity search

Sets the property dissimilarity threshold. Sets the threshold for similarity searches. Lower threshold results less hits which are more similar to the query structure MolSearch API


Not applicable.

JChemSearch API


        JChemSearchOptions searchOptions = new JChemSearchOptions(SearchConstants.SIMILARITY);
        searchOptions.setDissimilarityThreshold(threshold);
        // ...
        JChemSearch searcher = new JChemSearch();
        searcher.setSearchOptions(searchOptions);

Default value is 0.3.

JChem Oracle Cartridge


 Use the jc_compare operator with simThreshold:float.

The following SQL query returns the number of structures in nci_250k, whose similarity with Brc1ccccc1 is greater then 0.9:

SELECT count(*) FROM nci_250k WHERE jc_compare(smiles, 'Brc1ccccc1', 't:i simThreshold:0.9') = 1;

jcsearch command line tool


It can be set when the search type (similarity) is specified:

-t:i[:dissimilarity_threshold]