I can't believe no one has ever made a atomic-scale 3D model of a bacterium or a mitochondrion. At least I can't seem to find one. I totally want to ogle its molecules and see how they all fit together!

Shouldn't be surprising. A carbon atom is ~1 ångström in diameter. A mitochondrion (smaller of your examples) is 3 microns in diameter. That's 4 orders of magnitude in one dimension. Across three dimensions that's 12 orders of magnitude. Also, the atoms don't exactly "fit" together. Cells are fundamentally liquid, and the system has more in common with fluid dynamics than Legos. It's also a dynamic system so you need to answer the question of how long the model should simulate. Which adds another dimension and orders of magnitude of complexity. There's also the question of initial conditions. Since it's a dynamic system, a random initial configuration doesn't actually cover everything. Even the far simpler simulation of a 3-body solar system is complex enough that we're still finding new arrangements of stable orbits. And even after doing all this, a model wouldn't actually be good for very much. It would seem mostly like noise to us. Finding the right abstractions for the visualization would be far more difficult a problem. The visualization itself wouldn't be much help in finding the right abstractions, going by history. Lately I think of over-valuing visualizations as a cognitive bias. The visual cortex is evolved for a tiny, tightly circumscribed set of circumstances. We don't yet appreciate quite how far beyond its abilities we've gone.

I'm well aware that a simulation of an entire bacterium is generally beyond our computational abilities. To counterargue, 1. both bacteria and mitochondria vary a lot in size, and some are less than one micron in all dimensions. 2. a car is ~2 metres wides. The planet is ~20,000,000 metres from pole to pole. Across two dimensions that's (over) 14 orders of magnitude (510 trillion m^2). But individual cars are generally discernable in Google Maps. 3. I don't see what the problem is with the visualization. The person building the model knows the boundaries of each component and subcomponent (AFAIK the model cannot be produced automatically using, say, an electron microscope; experts in the detailed structure would have to make it, so they know the boundaries) so these boundaries can be easily highlighted for visualization. And with a GPU, local approximate simulations of parts of the cell should be possible. 4. I think lots of software developers feel a low need for visualizations compared to the general population, but I don't feel that way. It's weird to me to hear someone imply that we need not "visualize" a physical object.

Yeah, maybe it's not a bad idea. But seems reasonable to me that nobody's tried it yet. There has to be a focused economic incentive to do something on the scale of Google maps. (And we'll have to agree to disagree that Google Maps is a reasonable proxy for the complexity of the project.)

Like Drew Berry's animations?

Btw I found a reasonable-looking estimate of the number of atoms in a bacterium of E. coli weighing 10^-12 g: 93 billion, of which 61% are hydrogen atoms (and about 2/3 are water). It's unclear if this is a typical bacterial size but it does appear to be a typical E. coli size based on this page http://book.bionumbers.org/how-big-is-an-e-coli-cell-and-what-is-its-mass/

The BioVisions project at Harvard has mind-blowing videos around the web. Well worth watching. To think that each of our trillions of cells has 100,000 transport "workers" moving stuff around inside, is a mind-boggling concept.

Anything to help give people the proper sense of scale.