A reason to be excited about the next fifty years.

July 20, 2012 4:10 PM

Computer Simulates Full Organism for First Time "Maybe they'll computerize an entire human brain one day—or even just a couple of cells. For now, Stanford scientists have created the first-ever software simulation of a full single-cell organism, the New York Times reports."
posted by bookman117 (33 comments total)

If you tend, as I do, to read "organism" as "orgasm" you spend a lot of your time confused. This was still very confused, even by my standards.
posted by Bulgaroktonos at 4:24 PM on July 20, 2012 [32 favorites]

Let me know when they've added it to MAME.
posted by The Confessor at 4:24 PM on July 20, 2012 [4 favorites]

I would play the hell out out of SimAmoeba.
posted by Aquaman at 4:28 PM on July 20, 2012 [1 favorite]

Thank you Bulgaroktonos, I am not the only one.
posted by ambrosia at 4:30 PM on July 20, 2012

I also read this as orgasm at first. Pretty exciting either way.
posted by snofoam at 4:30 PM on July 20, 2012 [3 favorites]

Does anyone know where the original paper is? This is tantalizing but I'm still not totally sure what they mean when they say they've simulated a cell. OK, so they simulated all the genes, the RNA, the proteins, and the metabolites. At what level? Is it a deterministic model, that is to say are the interactions scripted? Or are they just putting it all in some kind of incredibly detailed biochemical simulator? How do they simulate the folding of proteins? Are they simulating every copy of every type of enzyme, or what?

I guess what I'm saying is that simulations are always simplified, stylized versions of whatever they are meant to model, and I don't know how simplified this particular model is.
posted by Scientist at 4:31 PM on July 20, 2012 [12 favorites]

I wonder if aspects of this could be gamified to help teach biology in high schools. One could set up the simulation rules a certain way and let kids work to a common purpose, like manufacturing and moving proteins to the right places in the cell, in order to keep it alive.
posted by Blazecock Pileon at 4:34 PM on July 20, 2012 [3 favorites]

My mind is also in the gutter, and my first thought at reading the "Computer Simulates Full Orgasm for the First Time", was "sure, but I bet it's not half as much fun".

Then I wanted to read Simulate as Stimulate...guess where that went.
posted by HuronBob at 4:34 PM on July 20, 2012

Oddly enough I read it as "stimulates first organism" and was even more confused than Bulgaroktonos.

This is pretty cool though. I'd to see their predictive accuracy.
posted by Tell Me No Lies at 4:34 PM on July 20, 2012

Oh, silly me. The paper was published in Cell, it says so right there in the New York Times article. Off to look at it!
posted by Scientist at 4:35 PM on July 20, 2012

Does anyone know where the original paper is?

The NYTimes piece links to the Cell abstract.
posted by Blazecock Pileon at 4:35 PM on July 20, 2012 [1 favorite]

“I find this fact completely fascinating, because I don’t know that anyone has ever asked how much data a living thing truly holds. We often think of the DNA as the storage medium, but clearly there is more to it than that.”

Yeah, we talked about this, um, maybe 2 years ago? Someone was claiming that you could just digitize DNA and wham, you had everything necessary to make a person.

This is just so, so, so not true, because the cellular mechanisms themselves are incredibly intricate. If the DNA is the program code, the cell mechanism is an unbelievably rich function library that's built right into your computer. If you don't duplicate most or all of that function library exactly, the DNA doesn't mean anything... and it's the nastiest, snarliest, most horribly designed library ever. And the DNA is doing all kinds of bizarre hacks like overwriting library functions at different times, and modifying itself. The complexity is hard to fathom for ordinary mortals. So, yes, DNA is 'just a program', but it's a program for a computer that's orders of magnitude better than anything we can presently build ourselves. However, we do appear to be getting to the point of starting to understand these biological computers, which is huge.

I really wonder how accurate this model is. If it's a good model, it'll be a major step forward... and I suppose you could argue that even a bad model would be progress of a sort. At the very least, it allows them to make hard, testable predictions about organism behavior, so it ought to improve our overall understanding very quickly, because even when it's wrong, it'll be giving us clues on how to be right.
posted by Malor at 4:36 PM on July 20, 2012 [6 favorites]

The Hacker News thread has some good info/links in the comments -- one of the team members (jkarr) is posting there.
posted by RobotVoodooPower at 4:36 PM on July 20, 2012 [5 favorites]

Oooh, thanks RobotVoodooPower!
posted by Scientist at 4:39 PM on July 20, 2012

OK, I think that the money quote as far as what I was trying to get at above is very nicely summed up at the beginning of the Hacker News thread RobotVoodooPower linked:

This is not (necessarily) a model of M. genitalium - it's a model of our understanding of M. genitalium, and as such incorporates in everything that current technology in biological sciences allows us to look at. It takes a huge amount of data from many sources and tries to bring it together on a scale not previously done. However, that data may have significant flaws, biases and ultimately considers only what we're good at looking at (technologically/scientifically speaking). It's awesome, and 100% the right direction for the field, but equally it is not a, "synthetic life form being simulated" as much as, "a very very complicated model which uses huge amounts of multidimensional data to try and replicate the behavior seen in that data".

That was my main question – whether they were somehow directly simulating the entire molecular complement of the cell, or whether they were only simulating the interactions between all of the processes that we know happen in the cell. It seems they were doing the latter, which is still a great accomplishment and may prove to be the start of an exciting new avenue of cellular research but which isn't quite what the New York Times was implying.

Very neat though! It is worth pointing out that what this lets us do is it lets us contrast our understanding of how a cell ought to work (based on the simulation) with our observations of how the actual cell truly behaves. That will help highlight new avenues of inquiry – show us our blind spots so to speak – and perhaps even give us some insight into what might be causing discrepancies between the simulation and reality. So it will help us improve and refine our understanding of the organisms we are simulating.
posted by Scientist at 4:46 PM on July 20, 2012 [7 favorites]

I really wonder how accurate this model is.

If it squawks, it's all good.
posted by localroger at 4:46 PM on July 20, 2012

Bad science reporting is bad. Who knew?

Interesting achievement, but clearly very short of what is claimed by the NYT.
posted by b1tr0t at 4:50 PM on July 20, 2012 [1 favorite]

For anyone following the Hacker News thread (which is excellent and highly recommended) the person from the team that did the simulation goes by jkarr and he's the first author on the paper.
posted by Scientist at 4:50 PM on July 20, 2012

Re the level of detail, jkarr posted this in the HN thread:

We used a combination of tracking concentrations and individual molecules. For specific macromolecules like DNA polymerase, RNA polymerase, ribosomes, etc we kept track of the position of each individual molecule. For other things like glucose, water, etc with very high copy number we just kept track of the copy number.

posted by hattifattener at 5:21 PM on July 20, 2012

If you prick it does it bleed?
posted by StickyCarpet at 5:45 PM on July 20, 2012 [1 favorite]

Wow, I started reading Accelerando a couple weeks ago, and this is like, the third scientific advance in the news that would bring us closer to that crazy future since I started reading.

For reference, an early plot point hinges on technology that allows lobsters to be uploaded into a computer by simulating them down to the atomic level.
posted by malapropist at 6:02 PM on July 20, 2012 [1 favorite]

Heh, just reading the headline, I thought to myself, "Mycoplasm, right? It's always mycoplasm." (I understand there are good reasons for that, I'm just amused I could name the model before opening the article.)

As far as the bad science reporting goes...

“These models are now in routine use around the world to study the metabolic properties of many organisms,” said Bernhard O. Palsson, a professor of bioengineering at the University of California, San Diego, who added that they were used commercially to formulate commodity chemicals and biofuels.

I am no idea what the fuck he is referring to here.
posted by maryr at 6:10 PM on July 20, 2012

From an admittedly cursory read, this is actually a lot more impressive than what I was expecting from the text above the fold. Markus Covert's group has been doing some really interesting work lately combining different approaches to modeling, like flux-balance analysis (pretty sure Covert's PhD was with Bernhard Palsson, who basically invented FBA -- which on preview, that quote is almost certainly about) and ODE models. It looks like this is their first try at a cell-scale unified model.

What's interesting to me is that it's not "just" a statistical description: it actually has some mechanistic modeling under the hood. But it also doesn't fall into the trap of trying to make a huge kinetic model that treats everything at the atomic level, either: for instance, fluxes through metabolism are determined by FBA, which is a type of constraints-based optimization and doesn't need any parameters about, e.g., enzyme kinetics. So they're able to use the best models for the individual subproblems they're interested in. This kind of multi-scale modeling is something lots of people have been talking about for a long time in the "systems bio" community, and while I definitely will admit this isn't really my field, it's by far the most impressive and complete example of that approach I've seen.

So yeah, pretty far from brains in a box, but still a real and impressive achievement and one I'm looking forward to reading more about.
posted by en forme de poire at 6:13 PM on July 20, 2012 [2 favorites]

running a simulation for the division of a single cell took half a gigabyte of data.

Half a gig! Now I'm just calculating how many tiny little robo-souls I could store on my thumb drive.
posted by Uppity Pigeon #2 at 6:37 PM on July 20, 2012

I think Scientist and en forme among others have the right idea. I don't have a lot of time to really dissect the article, but it seems really well done and a nice step forward. If I put it in terms of gaming, which the authors I'm sure would agree with, the NYT describes Skyrim, while in reality we are talking 8-bit atari. However, I personally think this is a huge step forward. To construct a biosystem will ultimately be far more instructive than the subtractive models we are stuck with now (knock out mice, etc.).
posted by oshburghor at 8:26 PM on July 20, 2012

...See, that's exactly where I think the in silico work has it wrong. I think it's great for narrowing down a list of possibilities or finding interactions we might no have predicted, but I think we'll be returning to knockouts and other wet lab technique to verify the in silico for a long while yet, if not always.
posted by maryr at 8:30 PM on July 20, 2012

The computer was probably faking it.
posted by horsemuth at 10:28 PM on July 20, 2012 [2 favorites]

malor: If the DNA is the program code, the cell mechanism is an unbelievably rich function library that's built right into your computer. If you don't duplicate most or all of that function library exactly, the DNA doesn't mean anything.

Thank you for a simple, elegant summary of what I've been trying to tell people for years.

BONUS: can also be applied to the 'selfish gene' perspective.
posted by lodurr at 6:55 AM on July 21, 2012

I wonder how realistic an environment for this Mycoplasma they can put together.

... because for a suitable, livable environment, we can run evolution. We can even tweak the constraints/environment until it outcompetes original Mycoplasma yet dies in contact with human skin and try to gene-splice it back into the real world. Then Mycoplasma goes away in five years of antibacterial soap.

With RepRap able to print everything but its circuit board, we can only think of the possibilities in genome-printing. It's been said that the 20th century was the Chemistry Age and the 21st would be the Biology Age.

Also there are things to be said about the Badiou v. Deleuze feud here, but they're so besides the point, the mean value theorem doesn't even apply anymore.
posted by syntaxfree at 7:00 AM on July 21, 2012

I work on the system described in Accelerando (although in crabs, not lobsters). It was more than a little jarring seeing my somewhat obscure area of research suddenly pop up in a novel I was reading. Unfortunately, we don't have any particularly good computer models of the STG (it's way more complicated than an amoeba). Furthermore, I'd put the odds of STG models achieving sentience at exactly zero. The STG has no role in cognition, it solely controls rhythmic movements in the stomach that enable crabs and lobsters to digest their food.

I guess that's a bit off-topic, but after years of banging away at trying to accurately simulate the activity of even one neuron in the STG, I feel like we're a LONG, LONG way away from brains on computers. AI may do amazing things soon (I have no idea, not my field), but if so, it won't be based on accurately simulating brains.
posted by Humanzee at 7:37 AM on July 21, 2012 [3 favorites]

Another jkarr comment on HN:

We used MATLAB. We organized the training data into a database (referred to as the "knowledge base") which was implemented in MYSQL with a web-based UI implemented in PHP. The various other scripts in shell and Perl were used to run the model on our Linux cluster.

Those are excellent tools for investigation. I wonder how much room there is for optimization, like porting the code to C++ and maybe using Postgres instead of MySQL. (Links are a bit of a joke with an underlying kernel of truth...)
posted by A dead Quaker at 10:06 AM on July 21, 2012

They should turn this into a game where the cell evolves and you have to fight other organisms for survival.
posted by pashdown at 12:17 PM on July 21, 2012

"A model of M. genitalium" does nothing to reduce the confusion over organism/orgasm...
posted by IAmBroom at 8:58 PM on July 22, 2012