Update JCIM workflows

[dominiquesydow]

Description

Update JCIM workflows (https://doi.org/10.1021/acs.jcim.6b00686); using KNIME 4.5.0.

Todos

Start updates for full workflows, then copy-paste changes to small workflows (which are a collapsed version of the full workflows thanks to meta-nodes).

See track changes below.

• GCRPDB_example (full/small)
• GPCR-kinase --- NOTE: Uses old KRIPO fragment library dataset
• Chemdb4VS (full/small) --- NOTE: Update small stand-alone node sequences if we can get some dummy dummy scores.txt and decoy.smi files
• KLIFS_example (full/small)

Not covered in this PR

• KRIPO_bioisosteric_replacement_workflow (full/small) --- FIXME: Installation fails for KripoDB nodes ("Fragment information" and "Similar fragments") with "Cannot complete the request. See the error log for details."; cannot see logs in KNIME Console
• SyGMa-example (full/small) --- FIXME: Using installed node "SyGMa Metabolites" fails with "The selected node could not be created due to the following reason: org/knime/python2/config/PythonCommandFlowVariableConfig"

Status

• Ready to go

[dominiquesydow]

Track changes for "GCRPDB_example"

• Replace “JSON to Table” node to fix problem with multiple ligands per structure
  • Problem: This node splits multiple ligands per structure in a single row by duplicating ligand-related columns (name, type, function).
    • This was not a problem with last execution 4 years ago, since all structures had only one ligand but now we have 3 structures with 2 ligands: 5D7D, 6N48, and 6OBA
    • Now, this causes problems with the next “Joiner” node, where we are joining structural data from the “GPCRdb Structures of a protein” node with IFPs from the “GPCRdb Structure-ligand interactions” node based on the PDB code and the ligand name (remember, we have currently multiple columns with ligand names)
  • Expected behaviour: We want to split multiple ligands per structure into multiple rows (thus, keeping only one column each for name, type, and function).
  • Solution: Replace the “JSON to Table” node with a sequence of
    • “JSON Path” - Set paths to the ligands' name, type, and function (columns are now called “Ligand name”, “Ligand type”, “Ligand function”)
    • “Ungroup nodes” - Split multiple ligands per structure into individual rows
    • “Column Filter” – Drop JSON column
    • Ideas taken from here and here (look for the posted json.knar file)
• Update deprecated “Joiner” node (GUI changed; transfer previous settings); FYI: Current execution results in a 1174x16 table

[dominiquesydow]

Track changes for "GPCR-kinase"

• NOTE: KRIPO fragment library based on PDB dataset 2016-06-29 (?)
• Update deprecated “Cell Splitter” and “Joiner” nodes
• FYI: Final table has 2268 entries (before 1576 entries)

[dominiquesydow]

Track changes for "Chemdb4VS"

• Replace broken ChEMBL node with Albert’s “ChEMBL – map UniProt to ChEMBL” meta-node
• Replace broken ChEMBL node with Albert’s “ChEMBL – bioactivities from target ID” meta-node
• Replace “calculate pAct” meta-node with “Extract pChEMBL values”, which extracts pChEMBL values for exact measures in nM; sets inactive compounds to 0
• Replace broken ChEMBL node with Albert’s “ChEMBL – compound information” meta-node
• Add new node: “RDKit Salt Stripper” to strip molecules off salts;
• Split molecules with bioactivity data into active/inactive at “Numeric Row Splitter”: For both sets, rename column “canonical_smiles” to “SmilesValue” and continue to work with the latter (renaming useful for RDKit).
• Added “Lipinski’s Rule-of-Five” node before Rule-of-5 filter (not sure if Ro5 was previously part of ChEMBL’s output)
• Added “Molecular Properties” node to calculate MW before MW>100 (not sure if MW was previously part of ChEMBL’s output)
• TODO: No dummy data available for stand-alone node sequences (thus not properly updated) that
  • reads docking scores (scores.txt) and plots ROC curves
  • reads in decoy SMILES (decoy.smi)
• NOTE: MarvinView needs license (ChemAxon)

[dominiquesydow]

Track changes for "KLIFS_example"

• Update deprecated nodes “Joiner”
• TODO: Could not test (and if needed update) the bar chart functionality, yet, due to missing R dependency; tried to use installed “Bar Chart” node but could not reproduce plot, so stick to R snippet

[dominiquesydow]

@AJK-dev could you please take a look if this PR is ok to go from your side? There are two workflows that I could not cover so far; to be covered in another PR maybe?

[blindpikachu]

with regards to the KRIPO_bioisosteric_replacement_workflow. Is there a script/workflow to create the kripodb from scratch? otherwise, how would you suggest to fix this workflow ?

[sverhoeven]

Steps to create the kripodb are at https://github.com/3D-e-Chem/kripodb/blob/master/docs/baseline-update.rst