3D Reconstruction of Human Hemoglobin

From molecular data to volumetric reconstruction

Human hemoglobin, scanned at 55 angstrom resolution and enlarged 18,549,827 times its original size to 4 inches.

So, about 6 months ago, I got introduced to David Gittleman, over at the McFarlane Design Group. Now, McFarlane Design Group is world famous for making some of the best action figures and toys produced today.  Todd McFarlane, himself a superhero comic book artist, entrepreneur, and author, just had his Amazing Spiderman 313 cover sell for 71K! David was interested in finding out more about converting laser scans of actors into textured models to be used in 3D animations and visual effects.  I did a bit of that a few years back, and we talked about that process a bit.  Curious to find out more about toy manufacturing, David and I also discussed some of the various aspects of their process.

Photo and Laser Scan of actor John Noble from Fringe)
http://en.wikipedia.org/wiki/3D_scanner
http://www.gamepro.com/article/news/215667/the-most-impressive-thing-i-saw-at-e3/

David, like myself, is a visual effects aficionado, and I’m a born again comic book geek, so we spoke the same language when it came to the creative process, and all good things sci-fi. But mainly we discussed ways that each other’s skill sets might find common ground. It turns out we had quite a bit. In both of our production processes, we take data sets from scanned objects and use them as building blocks for our final product.  In David’s case, its scanning actors and people to make toys.  And for myself, I find datasets of proteins and molecules to tell 3D animated stories.

In my line of work, whenever I tell stories that take place at a molecular level, I find proteins and drugs published on the web, and turn them into actors for scientific stories. Working with molecular data is certainly one of the most interesting aspects of modern medical animations. Reconstructions from this data provides us with an observable example of an unsee-able universe recreating proteins accurate at the atomic level.

Despite completely different factions of the computer graphics industry, what became clear from my conversation with David is that both types of production processes shared something in common.  We both rely on manipulating the scale of real world objects in order to create our 3D models. These guys take datasets, import them into a 3D program, and scale them DOWN to toy size.  I take molecular datasets, and scale them UP in order to tell 3D animated stories.  

“Just as an ocean wave striking a lighthouse produces secondary circular waves emanating from the lighthouse, so an X-ray striking an electron produces secondary spherical waves emanating from the electron.”http://en.wikipedia.org/wiki/Xray_crystallography

As David and I were wrapping up, he kindly offered their printing services “if I ever wanted to print something out.”  I don’t know if David knew what he was getting into, but the offer had me thinking.  I wanted to convert a molecular dataset into a printable mesh.  I’ve thought about this a number of different times since I started working molecular data as a medical illustrator.  By far my favorite protein is Hemoglobin.  Not only does it have an interesting conformational structure, it’s Heme Group  contains 4 iron atoms!  This is an organic protein, caring around a metal atom, using iron like a magnetic tool in order to transport oxygen throughout our bodies!  All of a sudden, here’s a chance to not only use an actual 3D printer, but have the help of some of the best toy makers out there.   David put me in touch with Michael Gulen, a true master of his craft.  Michael and I briefly discussed the specifics, and promised I send him a model just as soon as I could prepare one.

On my end, the process began with a good protein dataset of Hemoglobin.  A protein data file, at its simplest, is little more than a matrix of (x, y, x) coordinates with a little bit of additional information.  Each point represents the location of a specific type of atom, and is denoted with an atom type and sometimes a bit more about its amino acid group.  Combined together, you have a collection of all the locations of every atom in a protein.  Proteins range in size and shape. Some are very organized, and regular, some are random clouds of noise. Sometimes they’re as small as twenty or thirty atoms, sometimes they have tens of thousands.  For our experiment, I selected 1HHO.  Hemoglobin is a good size protein, at 4,659 atoms. Here you can see the first atom of the dataset is nitrogen (N), which is part of valine (Val) amino acid group.  The next three numbers denote Nitrogen’s (X, Y, Z) position.
From there, things get sent over to a 3D App for organization and object optimization.  Now, in hopes of extracting the most information possible from the data I’d collected as to create the most educational model possible, I wanted to give myself the option to create a mesh for each functional group.  The 4 major chains, 4 Heme Groups, and 4 O2 molecules.  But accessing that data, beyond just the surface structure of the protein, adds a layer of complexity to the project. In the images above, you can just see the edges of the Heme Group (white surface surrounded by grey surface), but the Iron atoms (contained within the white surface), and their bound 02 molecules (red) are on the inside of the protein. While all of the pieces of the protein fit together like puzzle pieces in their native state as data, we needed to ensure that am actual model would not fall apart once it came out of a 3D printer.  The challenge: how can we create a model that will both disassemble and reassemble, but when assembled, would structurally stay together. Thankfully, Michael enjoys a good creative challenge. We talked a lot about different possibilities: different ways to disassemble it, pinning structures in place with rods to keep things together, 3D rigid body simulations to test its structural integrity… But there was just no way to know if this was going to work. Once the printing process completed and we’ve cleaned up the model, what happens with to each piece once gravity gets ahold of it? Would it naturally hold itself together? And if it could naturally hold itself together, would that make it more likely to break? Could we even figure out how to put it back together if it did disassemble? That’s a lot of questions. The only way to resolve them was to test, and testing cost money. There’s nothing worse than spending a bunch of money and failing. Both Michael and I thought it best to make this an iterative process. We didn’t want to waste a print, we wanted results.  We wanted to produce a successful test, then make intelligent choices to refine that process for better specificity and functionality. Instead of sending 12 different pieces, serparated into functional units, we took a more modest first step. Let’s get an actual prototype manufactured, something we can hold in our hand, something tangible.
3D Print of Hemoglobin

3D Print of Hemoglobin

No comments yet.

Leave a Reply

%d bloggers like this: