KNIME workflows developed in the 3D-e-Chem project using KNIME nodes developed in the 3D-e-Chem project.
- Searching KLIFS for analogs of a ChEMBL library
- Chemical Diversity in the G Protein-Coupled Receptor Superfamily
- 3D-e-Chem: Structural Cheminformatics Workflows for Computer-Aided Drug Discovery
- Fig. 2: Structure-based bioactivity data mapping of kinase inhibitors
- Fig. 3: Scaffold replacements for kinase ligand design
- Fig. 4: Ligand-based cross-reactivity prediction
- Fig. 5: Sequence-based ligand repurposing within a protein family
- Fig. 6: Structure-based GPCR-kinase cross-reactivity prediction
- 3D-e-Chem-VM: Structural Cheminformatics Research Infrastructure in a Freely Available Virtual Machine paper
The 3D-e-Chem KNIME nodes are part of the
Community contributions. The
community contribution software site is disabled by default, it can be enabled
by in KNIME menu going to File>Preferences>Install/Update>Available software
sites and checking the checkbox of the Stable Community Contributions software
site.
The workflows can be imported into KNIME by importing each KNIME archive file (*.knwf). Before importing make sure the community contribution software site has been enabled, otherwise the 3D-e-Chem KNIME nodes will not be found automatically.
The Pymol session files (*.pse) can by opened with PyMol.
Below is the workflow for identifying the most similar ligand in KLIFS for a set of kinase inhibitors obtained from ChEMBL. This workflow is described in the main text and Figure S1 in KLIFS: an overhaul after the first 5 years of supporting kinase research in Nucleic Acids Research (doi:10.1093/nar/gkaa895).
This workflow collects and processes all known inhibitors for a given kinase from ChEMBL (based on a ChEMBL identifier according to the TeachOpenCADD workflows). As an example, the workflow collects all KDR (VEGFR2) inhibitors with a pIC50 value ≥ 5 from ChEMBL. This compound library is subsequently screened against all ligands in KLIFS using the ECFP-4 and MACCS fingerprints. The best hit for each compound, with a minimum Tanimoto similarity score of ≥ 0.4 for ECFP-4 and ≥ 0.8 for MACCS, is kept. Finally, a full list of all PDBs with this best reference compound is linked to each ChEMBL compound. In addition an interactive overview is generated for the remaining compounds for which no (close) analog was found in KLIFS.
Below is the workflow for the chemical diversity of GPCR ligands as shown in Figure 1 and Table 1 in Chemical Diversity in the G Protein-Coupled Receptor Superfamily (doi:10.1016/j.tips.2018.02.004).
This extensive workflow compares all co-crystallized GPCR ligands (taken from GPCRdb) to the known GPCR ligands (taken from ChEMBL) and assesses: i) the number of ligands of a specific crystallized GPCR similar to its co-crystallized ligands, ii) the number of ligands of a specific crystallized GPCR similar to co-crystallized ligands of any GPCR, iii) the number of ligands of other GPCRs that are similar to the co-crystallized ligands of a specific receptor, iv) the number of ligands of other GPCR subfamilies that are similar to the co-crystallized ligands of a specific receptor.
Below are the workflows shown in Figures 2 - 6 of 3D-e-Chem: Structural Cheminformatics Workflows for Computer-Aided Drug Discovery (doi:10.1002/cmdc.201700754).
Figure 2: Structure-based bioactivity data mapping workflow using KLIFS.
Figure 3: A workflow for the identification of potential scaffold replacements for kinase inhibitors while maintaining the protein-ligand interaction profile by combining protein-ligand interaction fingerprint (IFP) similarity with ligand-based dissimilarity (ECFP-4) analyses.
Figure 4: Ligand-based GPCR cross-reactivity prediction workflow that can also be used to derive ligand-based protein-protein (PP) assocations.
Figure 5: Workflow for the identification of ligand repurposing possibilities using a sequence-based double entropy analysis (ss-TEA).
Figure 6: A structure-based ligand repurposing workflow that searches for KripoDB pharmacophore similarities between GPCRs and kinases.
The initialization and combination of PLANTS KNIME nodes for docking runs requires great care. Therefore, an example docking workflow is available here:
3D-e-Chem-VM: Structural Cheminformatics Research Infrastructure in a Freely Available Virtual Machine paper
Below are the workflows mentioned in the 3D-e-Chem-VM: Structural Cheminformatics Research Infrastructure in a Freely Available Virtual Machine paper.
KNIME workflow archive files ending with _small.knwf are workflows with most nodes
collapsed into metanodes. The files ending with _full.knwf are workflows
mostly without metanodes.
Customizable KNIME workflow to extract GPCR ligands from ChEMBL and preparation for virtual screening.
Workflow to analyze the binding site similarity between all crystallized GPCRs and kinases. PDB IDs are fetched through GPCRdb and KLIFS nodes, and the KRIPO nodes are used to select similar binding pockets and to add ligand structures.
A Pymol session file called jcim/GPCR-kinase.pse is stored next to this workflow.
This example fetches the residues, structures, protein-ligand interactions, and mutations of the human beta-2 adrenoreceptor and its similarity to other beta-2 adrenoreceptors.
Data is fetched from http://gpcrdb.org website.
A Pymol session file called jcim/aminergic_alignment.pse with an aminergic alignment is stored next to this workflow.
Example for KLIFS nodes
This example performs interaction fingerprint analysis of human MAPK-ligand complexes.
Data is fetched from http://klifs.vu-compmedchem.nl/
Bioisosteric replacement workflow using Kripo Knime nodes.
A Pymol session file called jcim/KRIPO_3rze_2aot.pse with 3rze + 2aot Kripo fragment alignment is stored next to this workflow.
Example for the SyGMa metabolites node: predicting the metabolites of 5 (drug) molecules.