Not only proteins, but also lipids, sugars and nucleotides play roles in neurons, and eventually I will create ways of modeling these macromolecules in MAYA. Once a structure is represented three dimensionally in MAYA, a fourth dimension, time, can be added. Other molecules can be inserted into the scene and animated, the numbers and actions of which are limited only by what is known to science and published in databases. I have not yet developed a technique for explaining atomic bonds or chemical attraction in MAYA, but MAYA’s "Dynamics" menu set and the ability to write programs in MEL will allow me to do this in the future. MAYA does have the inherent ability to model simpler physical forces like gravity, magnetism, and diffusion.

Phase I will benefit when I update MAYA-Mol to mine more information from PDB files, specifically the side groups on the amino acids which make up the protein chains. Then I can assign charges to the atoms protruding from the side groups and use MAYA’s ability to model attractive and repulsive forces to create a van der Waals surface around each protein. Then, if informed which surfaces were supposed to stick together, MAYA could snap them into the lowest energy state and, therefore, the correct alignment on an atomic scale. But, as it stands right now, when I show proteins interacting they are just "eyeball" approximations of the appropriate binding activities, which do not demonstrate the chemical forces that make the actual proteins interact. And even MAYA-Mol learns science, it will only be able to recreate what real experiments have already determined about which proteins interact and with what surfaces – no program in the world can tell you that beforehand. Nevertheless, MAYA-Mol will be able to show the user what our data mean in four-dimensional space.

The movie below is an example of what this set of tools can do so far. The background is as follows. At the nerve terminal, vesicles filled with neurotransmitters wait for the electrical signal that means it is time to send an output to the next neuron(s). When that electrical signal arrives (animated as blue electricity), voltage-gated calcium channels (red) open briefly to allow calcium ions to diffuse into the nerve terminal where they are in low concentration from the outside, where they are in high concentration. Proteins on the vesicle detect this calcium and respond by changing shape, setting in motion a series of protein-protein interactions which cause the vesicle to fuse with the cell membrane. Fusion exposes the inside of the vesicle to the outside of the neuron, thus neurotransmitters are exported in 0.5 milliseconds. The proteins that interacted to effect this fusion are then transported to another site to be recycled. Only two of these proteins are shown. The light blue is syntaxin, a transmembrane receptor protein that positions the fusion machinery near calcium channels. The dark blue is synaptotagmin, which may be the primary "trigger" for fusion upon calcium binding and which recognizes syntaxin as a binding partner. According to X-ray crystallography, synaptotagmin binds 3 calcium ions as shown in the animation (Sutton et al, 1995).

There are three cautionary notes to be made. First, the precise three-dimensional structure of a calcium channel is unknown, so it is represented abstractly. Second, the published structures of synaptotagmin and syntaxin are incomplete, so they are only partially correct in the animation. And third, I did not animate the binding of these two proteins accurately for the reasons I mentioned in the methods section. On an optimistic note, the uncertainty whether synaptotagmin could be the primary calcium sensor is exactly the type of question I hope 3D models can ultimately help resolve, since we can actually watch it in action..

One problem with three-dimensional modeling is the massive amount of random access memory needed to model complex surfaces. Molecular surfaces are especially onerous, and when more than a few are animated, RAM becomes a limiting factor. MAYA is going to have an awful time calculating the electrical forces on a protein when I eventually depict them with side-chains, and may not be able to handle as many as I want even on a muscular computer. Another limitation I encountered was the lack of structural information for many of the key players in this system, like the calcium channel. Channel proteins cannot be visualized with current techniques like X-ray Crystallography and Nuclear Magnetic Resonance (NMR) because they are too big and live within the membranes of cells, so I may never have these structures to animate.

On the other hand, the simple test images I created this semester turned out to be more attention getting than I had anticipated. The explanatory value of this project will be unprecedented among free software available for molecular modeling -well, MAYA is far from free, but I will publish MAYA-Mol next year.

In conclusion, I believe these results support my first hypothesis, and this project is going to be valuable to me for organizing conceptual models and explaining where the novel proteins I investigate in the next several years fit into the molecular mechanics of neurobiology. Also, the resulting animations will add clarity, content, and character to my presentations. But I encountered limitations with MAYA that precluded any strong conclusions in favor of my second hypothesis. I do not foresee this project modeling these systems in the level of detail I initially expected. I plan to try other 3D software and investigate the potential for hybridizing MAYA-Mol to more accurate, specialized molecular modeling programs in the near future.