Fantastic Voyage Part 1: AutoFill on a synaptic vesicle with placeholder proteins and mock bilayer

• As of 2010, scientists can see cells, cell organelles, and even large macromolecular structures with microscopes. We can visualize molecules and a handful macromolecular structures with what I call 'nano-inference' technologies like xray crystallography and NMR.
• An algorithm we've developed called AutoFill, pools data from several sources to fill organelle, cell, and histological structure data gathered from microscopes (or computationally generated (a recursive scaling capability of AutoFill itself) ), with molecular detail including: structure data from the Protein Data Bank, the TransMembrane Protein DataBank, cryo EM databases, transcriptomes, translatomes (providing rough molecular weight for appropriate sized placeholder spheres), and many other databases.

HISTORY
• For over a decade, I've contemplated algorithms to help model the interface of these two resolution limits, aka the meso-scale that David Goodsell http://mgl.scripps.edu/people/goodsell has illustrated in watercolored drawings with meticulous research and comprehensive detail for two decades now.
• In 2004 I motivated to develop these algorithms. Lacking substantial programming education and experience, I moved in '05 to work next to David in Art Olson's Molecular Graphics Lab at the Scripps Research Institute to formally initiate this project as part of my PhD thesis. While working on other projects, I roughed together several crude versions of the the fundamental algorithms and received valuable feedback from my PI (Art Olson) and other programmers in the lab to help optimize aspects of my crude code.
• In 2008, we hired a talented high school intern, Mostafa Al-Alusi who worked with me to translate all of my old bits of code into a single object-oriented file with relatively clean style. Mostafa suffered through several of my tangential brainstorming sessions as we bounced ideas around and flowcharted our way through incredibly complex algorithm concepts. His coding ability, comfort with vector math, eagerness to learn/focus really helped push the code forward. He contributed several novel ideas and solo-scripted numerous helpful functions as well.
• In early 2010, Michel Sanner and I sat down to translate all of the code into a stand-alone python application... until this point, the algorithms had been locked in the obscure proprietary coding language of COFFEE- the native language of Cinema 4D in which I was most comfortable working. Michel, the PI behind Python Molecular Viewer (PMV) http://mgltools.scripps.edu/ is a masterful programmer and within a week and a half, we had the rough algorithm fully transposed and functioning along with several improvements that I hadn't yet had time to implement/debug. Michel and I worked out solutions to numerous limitations that had plagued the code and he quickly helped me work out a shift in the fundamental architecture that allows the code to finally work recursively and simultaneously on nested organelles without the tedious manual overrides that my original versions required. Michel has since continued to clean the code and to both simplify and augment it where ever possible. The new Python script is clean, properly modular, and clear enough for a Python noob like me to manipulate. It is structured in a manner that allows AutoFill to easily port new data and modeling algorithms as they become available/enhanced. Within a few weeks, we should have our 1st edition whole-cell fill generated. As indicated in this movie, Embedded PMV (ePMV) developed by Ludovic Autin with Michel, allows us to port Autofill back into a familiar professional animation environment like Cinema 4D to interact directly with the data and produce production quality imagery. I'll post and link more about ePMV in other movies.
• COMING TO AN ARCADE NEAR YOU: The 1/2 of an 80nm cube shown in this movie takes ~3 min. to fill on my 2007 mac. As anticipated, the Python implementation of the code slows the algorithm nearly 200 fold. Once we iron out other aspects of the code, we will translate the bottlenecks back into C++ and expect to be able to generate ~50x50x20nm cuboids of molecules on the fly, in real time and with Brownian Dynamics, e.g., providing a submarine style view of a cell's molecular details as indicated in my older COFFEE videos.
• Chris Arthur provided the gorgeous Synaptic Cleft tomogram for a cover image we slapped together for NeuroScience http://www.sciencedirect.com/science?_ob=PublicationURL&_tockey=%23TOC%23.... It shocks me every time how closely this data set resembles an illustration I created in 2004. http://sciencecareers.sciencemag.org/career_magazine/previous_issues/articles... (Nice support for the scattered research I pulled together for that old image.)

Category:

Science & Technology

Tags:

• Fantastic
• Voyage
• biology
• goodsell
• graham
• grahamj
• johnson
• mesoscale
• meso-scale
• molecular
• cellular
• autofill
• mgl
• olson
• graphics
• lab
• scripps
• the
• research
• insittute
• medical
• illustration
• animation
• synaptic
• cleft
• vesicle
• blood
• vessel
• Michel
• Sanner
• Ludovic
• Autin
• ePMV
• PMV
• Python
• Molecular
• Viewer
• Cinema
• 4D
• C4D
• Cinema4D
• Chris
• Arthur
• Ron
• Milligan
• CimBio
• kinesin
• myosin
• walking
• motor
• proteins

[leggoman3877]

great animation graham, just would say I couldn't read the white text at the bottom it conflicted with the background colours

[grahamj21]

@leggoman3877 Thanks leggoman. I fully agree, and doubly so with the compression on YouTube. I have a version lined up in After Effects with substantial drop shadows behind the text that should make it legible and keep it relatively quiet, but I'm waiting to fix a few other things along the way.

Thanks,

Graham

[Amotarzi]

This is really well done! Especially all those complex structures, the lipid membrane looks good, too, a few more interesting materials would give it a much more realistic appearance, but then this is more for demonstration purposes, well done!