molstar/molrender

Create macromolecular images

Usage

Molrender creates images of the following types: model, assembly, chain, chain-list, models, or all. Either as png or jpeg.

Create all images as PNG

molrender all /path/to/5cbg.cif /path/to/out/

Create all images as JPEG

molrender all /path/to/1nmr.cif /path/to/out/ --format jpeg

Building

Build:

npm install
npm run build

Build automatically on file save:

npm run watch

Build with debug mode enabled:

DEBUG=molstar npm run watch

Build for production:

NODE_ENV=production npm run build

Scripts installation

npm run build
npm install -g

Publish

Prerelease

npm version prerelease # asumes the current version ends with '-dev.X'
npm publish --tag next

Release

npm version 0.X.0 # provide valid semver string
npm publish

Contributing

Just open an issue or make a pull request. All contributions are welcome.

Funding

Funding sources include but are not limited to:

• RCSB PDB funding by a grant [DBI-1338415; PI: SK Burley] from the NSF, the NIH, and the US DoE
• PDBe, EMBL-EBI
• CEITEC

molrender's Issues

Q: Would it be possible to molrender an MVSJ file?

I am wondering how difficult it would be to render a molviewspec JSON file. This would be a really nice utility that would extend the value of the MVS system.

Assemblies from crystallographic symmetry are not colored with distinct colors

E.g. 1fs6

https://cdn.rcsb.org/images/structures/1fs6_assembly-1.jpeg

Is there a specific reason for assemblies not to be colored with the "Chain instance" preset instead?

Performance on Large Structures

• it's painfully slow to generate all images for some structures, 4v99 took almost 24 h, 3j3q and most ribosomes are fine however (which is why I didn't spot this sooner)
• the process gets slower over time, it's even worse if a large assembly gets rendered first
• looks to me as if some data is lingering, but GC/memory look normal to me and the canvas gets cleared after each rendering
• I'm able to run with comparable speeds to 0.4.1 when I create a fresh ImageRenderer instance for each chain - doing so doesn't result in a noticeable performance penalty but seems like a pretty heavy-handed fix

When profiling, most time is spent in drawElementsInstancedANGLE but to a way higher ratio with the current code.

0.5.0 @ ~5 min per chain:

 [JavaScript]:
   ticks  total  nonlib   name
  21221   60.1%   66.7%  LazyCompile: *drawElementsInstancedANGLE /Users/sebastian/IdeaProjects/molrender/node_modules/gl/src/javascript/extensions/angle-instanced-arrays.js:47:30
    503    1.4%    1.6%  LazyCompile: *query :161:15
    440    1.2%    1.4%  LazyCompile: *JPEGEncoder.encode /Users/sebastian/IdeaProjects/molrender/node_modules/jpeg-js/lib/encoder.js:614:25

 [Summary]:
   ticks  total  nonlib   name
  24985   70.8%   78.5%  JavaScript
   5199   14.7%   16.3%  C++
   1095    3.1%    3.4%  GC
   3491    9.9%          Shared libraries
   1625    4.6%          Unaccounted

Fresh ImageRenderer for each render operation @ ~0.3 s per chain:

 [JavaScript]:
   ticks  total  nonlib   name
   5445    3.5%    4.2%  LazyCompile: *drawElementsInstancedANGLE /Users/sebastian/IdeaProjects/molrender/node_modules/gl/src/javascript/extensions/angle-instanced-arrays.js:47:30
   3290    2.1%    2.6%  LazyCompile: *write /Users/sebastian/IdeaProjects/molrender/node_modules/glsl-tokenizer/index.js:92:17
   1661    1.1%    1.3%  LazyCompile: *query :161:15

 [Summary]:
   ticks  total  nonlib   name
  24613   15.9%   19.1%  JavaScript
  43122   27.9%   33.5%  C++
   9708    6.3%    7.5%  GC
  25941   16.8%          Shared libraries
  61115   39.5%          Unaccounted

@arose @dsehnal Do you have any idea what might be going on? Could this be another problem in headless-gl?