Scipion-EM-ProDy: A Graphical Interface for the ProDy Python Package within the Scipion Workflow Engine Enabling Integration of Databases, Simulations and Cryo-Electron Microscopy Image Processing

Abstract

Macromolecular assemblies, such as protein complexes, undergo continuous structural dynamics, including global reconfigurations critical for their function. Two fast analytical methods are widely used to study these global dynamics, namely elastic network model normal mode analysis and principal component analysis of ensembles of structures. These approaches have found wide use in various computational studies, driving the development of complex pipelines in several software packages. One common theme has been conformational sampling through hybrid simulations incorporating all-atom molecular dynamics and global modes of motion. However, wide functionality is only available for experienced programmers with limited capabilities for other users. We have, therefore, integrated one popular and extensively developed software for such analyses, the ProDy Python application programming interface, into the Scipion workflow engine. This enables a wider range of users to access a complete range of macromolecular dynamics pipelines beyond the core functionalities available in its command-line applications and the normal mode wizard in VMD. The new protocols and pipelines can be further expanded and integrated into larger workflows, together with other software packages for cryo-electron microscopy image analysis and molecular simulations. We present the resulting plugin, Scipion-EM-ProDy, in detail, highlighting the rich functionality made available by its development.

Keywords: cryo-electron microscopy; elastic network models; ensemble analysis; global protein dynamics; hybrid simulations; normal mode analysis; principal component analysis; software integration workflows.

Figure 1

Illustrative workflow using Flexibility Hub and Scipion-EM-ProDy to solve challenges with interpreting landscapes from continuous heterogeneity analysis of single particle images and obtaining reasonable structures. There are 4 main parts. (1) A general Flexibility Hub starts from a set of particles from traditional SPA; (2) A fast, flexible fitting pipeline based on Zernike3D creates several structures; (3) ProDy ensemble analysis generates a conformational landscape; and (4) The ProDy ClustENM (D) protocol can be used for improving atomic structures.

Figure 2

Overview of the Scipion-EM-ProDy plugin, with red arrows showing consecutive steps following user interaction. (a) List of protocols available within the plugin as shown on the left-hand side of a Scipion window. These are divided into categories with an order indicating how they should be used in pipelines and workflows. Double-clicking on one of these protocols yields a form as shown in panel (b). (b) An example protocol form is shown for atom selection. (c) Execution of protocols creates a workflow of inter-connected boxes, as shown. The workflow here illustrates importing of three D614G spike structures with selection of protein atoms (yellow) and ensemble construction and analysis via PCA (blue) and landscape projection (orange). (d) PCA results are shown with the ProDy mode viewer that opens up VMD and NMWiz (blue outlined box). The control window for NMWiz (left) has many options for visualising the structure with arrows and creating movies in the VMD structure viewer window (right). The average structure is coloured by the relative size of motion from most rigid in red to most mobile in blue. The green arrows show the first principal component where the two subunits slide past each other to close together (direction shown by arrows; towards 3-down) and open together (opposite direction; towards 2-up). (e) A projection plot from Matplotlib (orange outlined box) shows the conformational landscape from the ensemble in the space of the first two PCs, with each point being a structure.

Figure 3

Combined workflow for comparing existing experimental structures and simulations. (a) The workflow contains two main parts: experimental structure ensemble analysis (green shading, left) and ClustENM simulations (blue shading, right). The experimental structures are aligned into an ensemble (red box) and analysed by PCA (central blue box) and RMSD clustering (left orange box). Some representative clusters are also used for NMA (3 blue boxes outside green shading) to confirm the best structure for simulations via comparison of NMs and PCs (following green boxes). Some atomic structure tools (yellow) were also used including as a first step to prepare the structure for simulations. Coloured arrows connect to other panels showing their outputs. (b) A combined landscape is shown projecting the experimental ensemble in orange and the simulated one in blue in the space of the first two principal components from the experimental ensemble (axis values are in units of RMSD in Å). The main conformations are labelled, including the simulation starting structure (PDB: 7KEC). (c) An overlap matrix is shown comparing the first five non-zero ANM normal modes of the 1-up, 1-I structure (PDB: 7KEC) to the first five principal components of the ensemble with low overlaps in blue and high overlaps in green, orange and red.

Figure 4

Technical details of the ensemble building form and its usage. (a) The ensemble building form is shown with each of the drop-down menus expanded to show the options. The alternative parts of the form that result from selecting particular options are shown in yellow boxes connected to arrows selecting these options. (b) Selecting the DALI option for input structures adds new filtering options for the results from the web server. (c) Selecting to use another structure as a reference replaces the text box for the reference structure ID with a pointer selector for the new reference structure. (d) Selecting the custom option for chain matching brings up the custom chain matching dictionary generation wizards. The first wizard (circled in red) enables the replacement of entries in the dictionary by specifying a position and match order in the two boxes with red labels, as shown by the two red arrows. The second wizard (circled in black) recovers items from the dictionary to make it easier to check and update them. It fills the chain order and label boxes, as indicated by the black arrows. (e) The full custom match dictionary used in the analysis of the 24 D614G structures is shown, matching our previous study [65]. The first D614G structure determined from intact viruses (used as a reference here) is given chain order BAC and the others have orders BCA and ABC to match it. The rest of the structures have similar chain orders, giving two blocks with an exception for one of the structures (PDB: 7EB4). (f) The output of the ensemble building protocol for the 24 D614G structures is shown, highlighting the different objects generated and the summary information provided.

Figure 5

Illustration of GNM and mode comparison viewers. (a) The GNM viewer form is shown for the C$\alpha$ atom GNM from the workflow in panel (d), based on an open adenylate kinase structure (PDB: 4AKE, chain A; blue arrow going up). Certain results are highlighted by arrows and shown in panels (b,c). (b) The mode shape is plotted as a line with atoms moving in opposite directions have positive and negative signs. (c) The NMWiz viewer colours the structure by the mode shape. (d) An example workflow using GNM is shown using open adenylate kinase. Two different atoms selections in yellow precede two GNM calculations in blue using all atoms of chain A and only the C$\alpha$ atoms. The latter is used for dynamical domain decomposition (orange), and the two of them are used to illustrate mode editing (light orange) and comparison (green). (e) The output from the dynamical domain decomposition wizard using two modes (including the zero mode) is a structure coloured by dynamical domains in line with the behaviour in the first non-zero mode shown above. (f) The mode comparison viewer is shown for the comparison between the all-atom GNM and the extended CA GNM. (g) The result from selecting the option to display the row from comparing mode 7 (non-zero mode 1) from set 1 to all the others with cumulative overlaps enabled is shown. Correlation cosine overlaps are blue bars and the cumulative overlap is shown with a dark red line.