Schrödinger KNIME Extensions allow scientists to protoype, validate, automate, and deploy multi-step workflows. In this first installment of KNIME questions and answers, KNIME expert and Schrödinger applications scientist Jean-Christophe Mozziconacci talks about how to get started with KNIME, several workflows that can be used to prepare structures for rigid-receptor docking, and how to customize Schrödinger’s validated protocol for flexible-receptor (Induced Fit) docking.
--
Q: I’m new to KNIME, but like the idea of automating workflows and executing them with a single click. What are Schrödinger KNIME Extensions, and how can I get started?
A: The program KNIME allows users to execute sequential tasks by using the output from one “node” as the input to another. KNIME doesn’t actually run any of the modeling calculations – instead, when a Workflow reaches a Glide node, for example, that node will launch Glide and run the calculation as specified by the workflow, using input from a previous step.
In other words, by connecting two nodes in KNIME, you can use the output from one program as the input for another. For example, by connecting an appropriate MacroModel node to an appropriate Jaguar node, you could perform a molecular mechanics minimization using MacroModel and then refine the resulting structure by performing an ab inito optimization with Jaguar, all with one click (see Figure 1).
The complete set of Schrödinger-supplied nodes is called Schrödinger KNIME Extensions. However, instead of creating workflows from scratch, most users will get started by using or modifying one of the many existing workflows freely available at the KNIME Workflow Center.
Figure 1: An example of a very simple KNIME workflow that reads in a compound (Node 1), runs a force-field-based minimization (Node 2), performs a DFT optimization on the resulting structure (Node 3), and then displays the results in Maestro (Node 4).
--
Q: What are some good KNIME workflows to get started with?
A: The KNIME Workflow Center is divided into sections based on the tasks to be performed in the workflow. The following workflows can be used to execute common or useful tasks, and are a good starting point for modification and experimentation:
• Cheminformatics section: Substructure search. This workflow allows you to sketch a query structure and search for matches among target molecules.
• Pharmacophore modeling : Phase Shape screening.
Similar to the substructure search, this workflow allows you to read in a query structure from a file, and search a structure library for shape-based matches.
• Molecular mechanics: Compare conformation search methods. This workflow compares the results of different conformation search methods. It uses only one node per method – you might try adding new nodes that use different methods, or adding new nodes that use existing methods with different settings. At the end of the workflow, results are gathered in one table for easy comparison.
• Docking and post-processing: Protein preparation and Glide grid generation. This workflow allows you to align the binding sites of different PDB structures for the same (or similar) targets, automatically prepare the structures, and then generate Glide grids.
When you choose to download a KNIME workflow from our website, in many cases you’ll actually download several different workflows that are variants of the same basic task. The variants will have names like “Workflow 1,” “Workflow 1-2,” and “Workflow 1-3.” The higher the second number, the more complex the workflow, and the more it can do. Typically, additional complexity is introduced by “branching” a workflow – for example, “Workflow 1-2” would consist of everything in “Workflow 1” plus an extra branch that performs some additional function (see Figure 2). As you experiment with KNIME, try starting with the simplest version of a workflow, and then begin using more complex versions or modifying a simple version to create a more complex version.
Figure 2: Above, a simple KNIME workflow for clustering molecules by fingerprint. Below, a modified version of the workflow that uses a branch to return additional results.
--
Q: I need to generate the Glide grids for a set of PDB structures. What are the alternatives to preparing the structures and starting the grid generation manually?
A: The Protein Preparation and Glide Grid Generation workflow can be used to do this. This workflow uses the following nodes to perform the following tasks:
• Align Binding Sites: Align the binding site of each PDB structure
• Protein Preparation Wizard: Add hydrogen atoms, adjust protonation states, fill in missing residues, and more. This is an automatic process, but results should be inspected manually.
• Glide Grid Generation: Generate Glide grids around the ligands in each PDB file. If a ligand is not present, the center of the grid can be specified. The grid generation node also allows users to create H-bond constraints if necessary.
The Protein Preparation and Glide Grid Generation workflow has a branch that analyzes the composition of each protein-ligand complex, and it displays 2D renderings of the co-crystallized ligands.
--
Q: I’ve used the Maestro interface to run the validated protocol for Induced Fit Docking (IFD), but I would like to experiment with some variations on this procedure. Can this be done with KNIME?
A: Modifying and prototyping workflows is an excellent use of KNIME, and is in fact being used internally at Schrödinger to test possible future improvements to IFD, our flexible-receptor docking protocol.
The validated IFD protocol – the same one used by the IFD interface in Maestro – is available for download as a KNIME workflow from the KNIME Workflow Center. After running the calculation, this workflow uses the “Run Maestro” node to visualize the IFD results using a custom rendering scheme. For example, labels are used to make it clear which residues had their side chains truncated and then restored during the calculation. This custom scheme is applied using Maestro commands that are saved in the node settings – right-click the node and choose “Configure” to view these settings.
If you wish to modify the workflow or create your own, there is a Schrödinger node that will run all IFD steps in the validated protocol, and there are also separate nodes that can be used to run each step in the protocol independently.