?

Log in

No account? Create an account
Days of miracles & wonders: digital viruses - Synchronicity swirls and other foolishness

> Recent Entries
> Archive
> Friends
> Profile
> my rpg writing site

November 18th, 2008


Previous Entry Share Next Entry
01:38 pm - Days of miracles & wonders: digital viruses
In reading pazi_ashfeather's lj, I encountered a link to a story that is likely the most shocking and impressive that I've seen in the last year or so:
Using a real-life virus as a model, researchers have built a virtual version using more than a million digital atoms.

Scientists have previously simulated small pieces of living cells, but researchers say this is the first digital simulation of an entire life form.
In addition to the obvious fact that being able to do this routinely provides an ability to create and test new antiviral treatments and thus vastly improved medicine, running this through Moore's Law, we get the ability to simulate a human cell (or other vertebrate cell) in approximately 30 years, and bacteria will be well before that. This discovery also once again makes me wonder about the idea of simulating an entire universe down to the atoms and if we are living in such a place (and if there is then any difference between that sort of universe and the "natural" universe we mostly imagine that we live in. I suppose the most notable difference is that in a digital simulation there actually is a creator or group of creators.
Current Mood: impressedimpressed

(12 comments | Leave a comment)

Comments:


[User Picture]
From:onyxrising
Date:November 19th, 2008 08:48 am (UTC)
(Link)
I miss you. Dinner? Saturday?
[User Picture]
From:onyxrising
Date:November 19th, 2008 06:13 pm (UTC)
(Link)
Nevermind. Got claimed by Elly.
[User Picture]
From:alephnul
Date:November 19th, 2008 10:01 am (UTC)
(Link)
RE: Universe simulations, You, of course, read xkcd?

As pazi_ashfeather points out, you are neglecting the time scale in your Moore's Law extrapolation. I'd add to that that we know the structure of viri at the atomic level, while we have no clue of the structure of a fruit fly at the atomic level, much less of a human body, nor do we have any imaginable method of getting the atomic structure of a living fruit fly.

Still, pretty cool.
[User Picture]
From:heron61
Date:November 19th, 2008 10:11 am (UTC)
(Link)
Learning the atomic structure of a single cell sounds possible (if admittedly quite difficult), and once you have that for all of the types of cells involved in a particular organism, then I think you're well on your way to simulating the organism.
[User Picture]
From:alephnul
Date:November 19th, 2008 10:39 am (UTC)
(Link)
You can't crystalize a cell, so you can't learn its atomic structure by x-ray difraction. You can learn the atomic structure of a protein that way, but a typical cell contains at least 2000 proteins at any given time (if I'm reading this right), plus plenty of other substances. The total human proteome is estimated at a million proteins even before you start counting post-transcription modifications to those proteins (only 20-30,000 genes, but ~80% of genes have alternate splicing patterns).
[User Picture]
From:alephnul
Date:November 19th, 2008 10:46 am (UTC)
(Link)
Plus, you'd need to work out the folding patterns of each of the proteins, and any alternate folding patterns.

And none of it is static.

My bet would be (accepting Moore's law continuing unlimitedly into the future) atomic scale modeling of a bacteria within the next 30 years, modeling of a yeast within 60 years, modeling of a fruit fly within 100 years, and modeling of a human within 200 years. Add 20 to 30 years at least to go from an idealized representation of each of those to actually being able to model one of those in a moderately complex situation with reasonable accuracy. Say, in 300 years (with forever continuing Moore's law) I could imagine atomic scale modeling of a human being from conception to death from old age.
[User Picture]
From:alephnul
Date:November 19th, 2008 10:29 am (UTC)
(Link)
I'm trying to imagine crystallizing a fruit fly. I'm not even going to try to imagine the computational difficulty of modeling x-ray diffraction on a fruit fly crystal.

Also, aren't you off by several orders of magnitude on whatever scale you choose to use? Moore's law, assuming that it has no limits, would give a speed increase of 32,000 x over the next 30 years. In what dimension is a human only 32,000 times as complicated as a virus? The modeled virus was 1 million atoms, while a person is 7 x 10^27, 10x21 times more atoms than a virus (which makes sense, as viri have masses at the 100s of attograms scale, 10^-21 g), and I would expect the difficulty of modeling to scale at a rate worse than linear, since what needs to be modeled is the interactions, not the atoms as isolated objects.

I calculate Moore's Law to require 144 years to get computers to be 4 x 10^21 more powerful, although I seriously doubt that Moore's law holds true over that many orders of magnitude.
[User Picture]
From:heron61
Date:November 19th, 2008 10:47 am (UTC)
(Link)
Ah, what we have is serious punctuation fail on my part - what I meant to say was "human, or other vertebrate, cell", but when I'm not being paid to type in actual English, my impulese is to use far too many parenthesis, and the result was massively unclear. A single human cell has (on average) a bit over a trillion atoms. 30 years (if Moore's Law doesn't fail before that) gives a 2^20, or approximately 1 million-fold increase in speed and memory, thus allowing a single human cell to be simulated. Simulating a human fully involves simulating between 10 & 100 trillion cells (a quick google ran into this range) which would take another 60 years after that (except that I'm betting on Moore's Law failing due to hard limits somewhere between 25 and 50 years, but the world will be a more fascinating place if it doesn't).
[User Picture]
From:alephnul
Date:November 19th, 2008 10:54 am (UTC)
(Link)
Okay, that makes much more sense.

Yeah, assuming Moore's law lasts for another million fold increase (and I'd bet it does just based on the rule of 5% and 95%, or its even simpler cousin, the rule of 50%; that is, something that has continued for X years will probably continue for somewhere between 1/20 X to 20 X more years. Moore's law has held for 50 years, so it should hold for between 2 1/2 more years and 1000 years, with a fair chance that it last another 50 years), I could see atomic scale modeling of a single cell being possible within this century.
[User Picture]
From:xuenay
Date:November 19th, 2008 10:58 am (UTC)
(Link)
Of course, once we have simulated individual cells in atomic detail, we'll likely have enough information to abstract their workings into higher-level rules without needing to simulate every atom in the whole-body sim.
[User Picture]
From:alephnul
Date:November 19th, 2008 11:33 am (UTC)
(Link)
I think we'll be able to abstract the workings of cells sufficient to run sub-cellular scale based models of full organisms before we ever reach the point of being able to do atomic models of cells. An atomic model of a cell will probably tell us stuff about cell operation that we won't be able to learn directly from working with cells, but those things will mostly not be sufficently critical to be able to do a reasonable sub-cellular scale model of a multi-celled organism.

We can model an ocean, and we can model a tea cup of water. The ocean scale model models everything down to km scale processes, and then abstracts the rest, while the tea cup scale model goes to molecular scale. Admittedly, in that case it is almost purely a matter of computational power that keeps us from doing molecule scale models of oceans.
[User Picture]
From:xuenay
Date:November 19th, 2008 11:38 am (UTC)
(Link)
True enough.

> Go to Top
LiveJournal.com