Use Case C, Worflow 3

Workflow used to obtain data for the research paper ‘The application of the Open Pharmacological Concepts Triple Store (Open PHACTS) to support Drug Discovery Research’, currently submitted to PLOS ONE.  

Requirements:
- Knime v2.9
- Open PHACTS Knime nodes 1.0.0 (DON'T use any later version!): https://github.com/openphacts/OPS-Knime

Installation:
- Download "org.openphacts.utils.json_1.0.0.zip" and unzip it in the plugins folder of your KNIME installation
- Download the workflow (Use Case C_Workflow 3.zip)
- Start your Knime environment (you should see already a couple of new nodes at the ‘Node Repository’, like OPS_KNIME)
- Import the workflow (file>>import-workflow...>>select archive file)

Running:
- Double click on the desired workflow
- Reset the OPS nodes (in orange)
- Execute the nodes. Because of a bug found, the OPS nodes have to be run one by one to avoid the retrieved columns of the metanodes get mixed up.

Description of the Workflow:
1. We start with targets in the form of text and obtain concept wiki urls of all homologues for desired targets using 'Free text to Concept for Semantic Tag' API call.
2. With the list of conceptwiki ID from above, we use the ‘Target Pharmacology: List’ API call to retrieve pharmacology data from ChEMBL. In this case we return the data using ‘_pageSize=250’ in combination with a loop over all result sets with the ‘_page’ parameter. Then we apply filters (2.1) to remove: Unspecified Activity value, Activity comment = inactive, inconclusive, Not determined and/or Not active and Potential data errors (values larger than 1E08 are removed) using ‘Row filter’ and ‘Row Splitter’ nodes. In this question we have also removed those compounds where in the "assay description" field appears the words "Inhibitor" or "antagonist".
3. We get compounds bioactivities only for compounds hitting more than one target. To do this we have used 3 nodes: ‘Pivoting’ node: we have grouped rows by ‘Compound name’ and pivoted by ‘Target name’ getting the minimum value for each target/compound.; ‘Column aggregator’ node: we count how many times a compound hit a target and ‘Numeric Row Splitter’ note: we split those compounds hitting more than one target.
4. In order to obtain binding affinity assay data for bovine, porcine and human VDR and DBP. We apply filters to limit data to desired species (Homo sapiens, Bos taurus, Sus scrofa) using the ‘Nominal Value Row Filter’ node. We include the rows with ‘binding’ and ‘affinity’ in assay description column by using the ‘Row filter’ node, concatenate the results and remove duplicates.