Tom Goddard
March 27, 2014
Here are some features of the Chimera 2 desktop application Hydra that are either not available or are much improved over the current production Chimera 1.
We would like to have a Chimera 2 release as soon as possible, but it needs to have new features that make some researchers want to use it. What should those features be? Some ideas below.
|
|
|
|
|
Features of Chimera 1 not in Hydra are too numerous to list, but the Missing table column is a sample.
| Category | Existing Capability | Missing Capability |
|---|---|---|
| Platform | Mac, Linux. Development on Mac, infrequent Linux build | No Windows build. |
| Data types | Density maps (mrc, ccp4, hdf5, brix), Molecules (pdb, mmcif), Surfaces (collada, stl), Sessions (.hy), CellPack (*.apr, *.xml), Traced Neurons (*.swc). | No sequences, no mol2, only most common denisty map formats. |
| Databases | Fetch from PDB or EMDB database with local caching. | No PubChem, ModBase, UniProt, EDS, .... |
| Display styles | Map display in all styles, surface, mesh, grayscale, planes, ortho planes, box. Molecule display stick, ball/stick, sphere, tube, surface, biological unit. Color by element, chain. | No ribbons (only tube), no wire, no nucleotides, no pipes and planks, no text labels. |
| User interface | Single window Qt user interface. Scenes. Recent file thumbnails. Open model thumbnails. Status line, reply log output, command history. Selection only of whole models. | No dialog interfaces, except open, save, map fit list. No selection and action menus. Can only select full model. |
| Input devices | Keyboard, mouse, trackpad, multitouch, menus, shortcuts, toolbar and command input. Space navigator, oculus rift, leap motion. | |
| Commands | open, save, close, scene, volume, fitmap, molmap, show, hide, align, surface, windowsize, camera, lighting, device. Atom specifiers, model, chain, residue, atom, residue ranges, negation, no zones, no bool ops & |, no ligand/protein/solvent... | No analysis commands, no movement commands (turn, rock), no movie recording, no focus, no coloring command, .... |
| Graphics | Single layer transparency. Mouse over pop-ups to identify atoms. All new opengl graphics (OpenGL 3.3) in Python (PyOpenGL). Two lights. Selection outlines. Sequential stereo rendering. | No silhouette edges, no shadows, no multilayer transparency. |
| Analysis | Fitting molecules or maps in maps. | Nothing for analysis. Can't edit or delete parts of molecule. No molecule analysis other than alignment (match). No distances, no hbonds, no morphs, no MD, no hydrophobicity coloring, no add charge, no add hydrogens... No sequence capabilities of any kind, no blast search, no seq-based structure alignment (MatchMaker). No web services. No density map processing, filtering, masking, morphing (vop command). |
| Output | Save images, any size, no supersampling. Save sessions including scenes (links to pdb, map files and database fetches). Save density maps. | Can't write PDB or mmCIF. Can't export scene. |
| Documentation | Html manual shown in app. Programming API documentation (Sphinx). | Minimal user documentation, partial programming API documentation. |
| Infrastructure | Python 3, OpenGL 3.3, Qt 5. Python hydra package needs no environment variable setup. | No saved preferences. No nightly builds or downloads. No test suite. No task manager. |
| 
| 
| 
|
| Fast PDB loading, displaying and memory efficient, 30x faster, 10x less memory than Chimera 1. | Solvent excluded molecular surface calculation that works for large molecules. | Solvent accessible surface area calculation that works for large molecules. | Structurally align hundreds of models with one command. |
| 
| 
| 
|
| Example sessions shown for new users | Recent file thumbnails, click to open. | Session and exported maps have file icon image of data on Mac. | Fast selection outlines, single pass shader rendering. |
| 
| 
| 
|
| Model list with thumbnail images to show/hide models. | Scenes handle everything sessions handle. | Interactive cross-fade when switching scenes. | Interactive motion blur. |
| 
| 
| 
|
| Sessions contain no executable code, safer to distribute on web, better forward compatibility. | Oculus Rift stereo goggle support. | Reply log can have images and links, uses html. | Cell pack reader handles collada surface files and xml files. |
1. Scenes - fast, complete, streamlined user interface, smart interpolation methods
2. Comparisons of hundreds of models - blast to find models, align, show mutations, insertions, deletions, different ligands, conformations, different binding partners in summary structure view with ability to quickly locate and compare structures of interest.
Half of PDB protein chains belong are similar to 60 or more other PDB chains at 30% sequence identity. PDB sequence clusters at 30% identity.
3. Multi-domain assembly, Sam Hertig's project, assemble tens to hundreds of PDB models for a sequence.
4. Multi-molecule assembly, use hundreds of available structures to model complexes.
5. Large model handling, million atom mmCIF models.
6. CellPack, build assemblies of thousands of proteins (e.g. whole HIV virus) from components using statistical placement on and inside surfaces.
7. IMP / RMF setup and view results for calculations of large assemblies like a nuclear pore with experimental restraints, ensemble analysis.
8. Segmentation. Researchers want to paint colors in 3d in a flexible way. Also want reproducible methods to paint objects in crowded density maps. This would take much research to figure out what to implement.
9. Map time series. This would be for 3-d live optical microscopy like Dyche Mullins cells moving in collagen. It is done poorly in Chimera 1. It is far removed from molecules, but is a new software area being opened up by fast 3-d microscopy.
10. Fitting models in high resolution EM maps (3-4 Angstroms). There is a lot of excitement in EM attaining near x-ray resolutions. Maybe use Rosetta fitting. Maybe a tool to show where residues don't fit well. Color by local resolution of map.
Scenes (1) and Structure Comparison (2) have the widest base of users who would be enthusiastic for a high quality implementation.
Scenes alone wouldn't be sufficient reason to use the program.
Assembly of multi-domain protens (3) or complexes (4) are attractive for a narrower group of researchers, and are a natural extension of comparing large numbers of structures (2).
Very large assembly tools (mmCIF, CellPack, IMP/RMF) are the most exotic leading edge stuff, with fewer researchers.
Structure comparision and assembly and large mmCIF models (2,3,4,5) leverage higher performance molecule handling.
Density map ideas (segmentation, time series, high res fitting) are not fleshed out enough and have rather limited user base.