Indigo-pains-recursive

The current Indigo library (1.1.13) cannot take advantage of the parallelizing nodes in KNIME the way the RDKit library can. One way to speed up the PAINS workflow when using the Indigo library would be to use a recursive loop that constantly shrinks the list of structures being presented to the next PAINS query as structures are knocked out of the main list.

Using the WEHI-10k reference set (embedded in the workflow), I get ~10% time reduction. Not as good as parallelizing the process, but welcome nonetheless.

---
2.80GHz single CPU
KNIME 2.8.0
Indigo nodes 1.1.13.201306111153

        time, PAINS/OK
multimatch 33m04s, 697/9303
recursive  29m30s, 697/9303

saving: 10.8%
---

At the end of the process, the list of failures indicates only the first PAINS query that was matched, rather than multiple PAINS queries that were matched to that structure.

Otherwise, exactly the same number of hits (and structures) are obtained.

See 'Attributions' over on the right for other versions of this workflow.

If you like this workflow, please reference our paper doi:10.1002/minf.201100076.

The Indigo nodes require installation of the 'community nodes' in KNIME. See http://tech.knime.org/community and http://ggasoftware.com/opensource/indigo/knime

Based on simple sub-structure workflow by James Davidson
http://tech.knime.org/forum/rdkit/substructure-search-with-rdkit
and recursive-loop filter workflow by Iris
http://tech.knime.org/forum/knime-users/row-filter-with-mulitple-criteria-question#post-27527

Modified to use SD/SMILES file as input, or manual SMILES entry, or database search.