Skip to content

Latest commit

 

History

History
122 lines (102 loc) · 7.27 KB

README_ENH.md

File metadata and controls

122 lines (102 loc) · 7.27 KB
Raw

a-ngl

Prototype: WebVR implementation of molecular structure viewer NGL. Ethan's README section.

To do (remote dev):

  • Check if another component on molecule entity (new mol-physics component) can access shape_* component coordinates and radius
  • intermol A-frame component
    • Electrostatic/hydrophobic in schema
    • Long term goal: remove physics entities and build CANNON.js direclty into ngl-mol component
  • Electrostatic interaction (CANNON.js level)
    • Done for now. Works pretty well with stiffness=20, damping=100
  • Finish updating a-ngl.js for "wrapper-less" physics body
    • Have addVoxelColliders store voxel radius and coords for addInterface
  • More detailed/accurate sphere voxels
    • Voxel centers only at relevant/sidechain atoms
    • Variable radius
    • Cylinders/multi-sphere for long residues like ARG

To do (with headset):

Change Log

ENH 4/7/2019

  • Representation algebra changes now use NGL handlers to add and remove representations, instead of reloading the entire stage

ENH 5/2/2019

  • Bounding box changed to bounding sphere. Radius set to smallest of length, width, or height of the NGL scene bounding box

ENH 5/3/2019

  • Bounding sphere shrinks to current representations instead of the entire structure.

ENH 5/6/2019

  • Started working on putting in residue interfaces. My game plan is to have NGL find residue centers and radii, then transform those representative spheres into the A-frame scene as a-spheres
  • Then add A-frame level CANNON.js interactions on spheres as separate component.
  • Working on building the basic framework today. Residues selected as interface residues are indicated in the ngl-mol schema as interfaceAlgebra parameter. Getting the tranformation from NGL to A-frame coordinate space figured out as well.

ENH 5/8/2019

  • Removed wrapper entity entirely. The physics body that represents the structure is now defined using a compound body composed of many sphere__* components. Tested compatibility with superhands with multiple molecules. I was able to very intuitively grab two molecules, throw/bounce them (as before), and actually "dock" the proteins (sterics only, of course). I will now start to work on getting the a-ngl component cleaned up to correctly handle attribute changes. For instance, changing molecules does not yet work. I will consider removing the change PDB code altogether, since it might not be that relevant (user would probably just want to delete a molecule entity then add a new one).

ENH 5/10/2019

  • Started working on electrostatic interactions in A-frame. Forked the spring interaction from A-frame physics, which is basically an abstraction of CANNON.Spring. So I also forked CANNON.Spring and changed the applied force from a harmonic oscillator to an electrostatic interaction (force goes by -k/r^2). Works pretty well, but I still need to remove all the attributes in CANNON and A-frame that are not necessary for an Electrostatic interaction, such as damping. I will also probably want to add multiple targets to the Electrostatic component, which I might rename Molecular so that I could include electrostatic/hydrophobic/VDW interactions as a schema parameter. First things first, I want to get CANNON.Electro working well:
    • Electro is not "on" until bodies are close enough
    • Electro is ignored at restLength
    • Coulomb's constant is replaced with the schema parameter for spring stiffness, k, at least for debugging purposes. k=-20 works pretty well for dynamic body of mass 1 (kg?)

ENH 5/15/2019

  • Electrostatic (or pseudo-electrostatic) CANNON.js physics work pretty well, and are currently tied to scene via intermol A-frame component placed on a single entity. I will next re-write the intermol component to handle multiple CANNON.js interactions so that it can handle all the intermolecular (electrostatic & hydrophobic) forces on the molecule entity.
  • ngl-mol will detect when separate molecules collide. Then, ngl-mol emits addElectroBody events to mol-physics.
  • I could also move ngl-mol.addVoxelCollidersByRes to the new mol-physics component, so that all physics are handled separate from ngl-mol. Still debating whether I should do this or keep it in the ngl-mol component.
    • Or, sequester physics bodies into a separate child or parent entity. This would allow for removal of shape_* components via removal of the entire entity.
  • The mol-physics component outline:
    • schema
    • init
      • Add event listeners for addElectroBody and removeElectroBody

ENH 5/16/2019

  • Got surface representations working. Need to set the representation component as an attribute of component instead of temp variable "r"

ENH 5/23/2019

  • Did a lot of work on the core ngl-mol component:
  • this.repAlgebra is now a dictionary/Object, not a 2D array. Much easier to manage.
    • repAlgebra as defined in inline HTML will now convert to an instance of NGL-type RepresentationParameters
    • For instance, representation specific options such as useWorker can now be defined in HTML inline
  • Squashed some of the bugs associated with surface representations. Sometimes, surface will not appear on initialization (see known issue entry from 5/22), but should show up properly with dynamic repAlgebra changes via setAttribute, so this is probably not a big deal.
  • Added the addRebuildListener() method that listens for rebuildNglRepsEvt events and responds by calling structComp.rebuildRepresentations()

ENH 5/24/2019

  • Added support for distance representations. In repAlgebra, simply define selection algebra for atom1 and atom2 (same syntax as sele parameter) to draw distance between exactly 2 atoms. As always, user has full access to all options in DistanceRepresentationParameter from HTML inline (e.g. labelUnit, labelColor, labelFont, etc.)

Known Issues

ENH 5/3/2019

  • When page loads, white cube appears and portions of the protein do not render correctly. When using dev keybinds, proteins usually load correctly, but sometimes these weird white cubes appear.
    • More graphically intense representations tend to trigger this more frequently
    • Calling structComp.rebuildRepresentations() after page load fixes this most of the time, but will still trigger occasionally
  • Also, shrinkWrapper does not seem to set Object3D correctly on first load
    • No longer a method. Resolved

ENH 5/6/2019

  • The wrapper entity physics body is defined by the HTML inline geometry, not by the wrapper geometry defined in ngl-mol component
    • Wrapper entity is deprecated. Resolved.

ENH 5/8/2019

  • Physics components do not have update methods (or at least ones that work correctly). el.setAttribute cannot alter, but will correctly add, physics components. In other words, do not define physics components in HTML inline, but rather through the ngl-mol component. Alternatively, use el.removeAttribute to completely remove the component before re-adding it.

ENH 5/9/2019

  • Physics shape components apparently cannot be removed currently (per console error from shape component). I'll have to find another way to update/remove shape components

ENH 5/22/2019

  • Sometimes surface does not show up. If I call sc.addRepresentation later, this happens less often, making me think it needs to wait until the stage (or some other part of NGL) is done loading before rendering. I should note that NGL will usually not throw an error for this.