Open-source biotechnology

Posted Mar 31, 2010 16:10 UTC (Wed) by nix (subscriber, #2304)
Parent article: Open-source biotechnology

A nucleotide is able to encode two bits rather than one, and the underlying processor is smaller, wetter, and smellier, but it's a program nonetheless.

So the protein folding problem simply doesn't exist? Oh, and programs on (silicon) computers generally have every bit of the program able to modify every other bit, and they're generally based on statistical mechanics and solute chemistry, and exist in an environment in which a vast supply of enormously older and more capable systems are actively trying to eat them.

I guarantee that any molecular biologists in the readership are having trouble restraining hoots of hysterical laughter. Such naivete: the comedown will be hard.

Open-source biotechnology

Posted Mar 31, 2010 16:39 UTC (Wed) by flewellyn (subscriber, #5047) [Link] (8 responses)

Given that protein folding is a problem for, well, protein, and not DNA...

Open-source biotechnology

Posted Apr 1, 2010 6:49 UTC (Thu) by mitchskin (subscriber, #32405) [Link] (7 responses)

A lot of DNA expresses itself through proteins. The point nix is making is that a biological system is much, much more difficult to engineer than a software system. The protein folding problem is part of it; that problem makes it very difficult to predict what your DNA modifications will do (you have to solve an n-body problem with a very large n just to figure out what shape your protein will have, and that doesn't necessarily tell you very much about how it will behave).

A big part of engineering (certainly software engineering) involves encapsulation: the hiding of complex behavior behind a simple interface. But in biological systems this is very hard to achieve; there are a lot of processes going on all mixed together in the same solution. That's what nix meant when he sarcastically described "every bit of the program able to modify every other bit"; of course we don't program that way because doing software that way would be total madness. But in biological systems, that's the way things are.

So (to restate what nix was saying): Imagine trying to program a computer system where 1. the behavior of the system is not at all deterministic, 2. the system prevents encapsulation, 3. you can't predict what effects your changes will have.

Evolution has had billions of years on Massively, Massively, Massively parallel hardware to come up with the complex systems we have today. Humans are now only barely able to scratch the surface of the complexity of those systems. There's a lot of excitement now about engineering biological systems, but it's a very long-term project.

Point being, be very careful about making facile analogies between biological systems and computer systems. There are some similarities but there are also very, very significant differences.

Re: the point about other systems trying to eat yours - I actually think internet-connected computer systems are kind of like biological systems that way.

Open-source biotechnology

Posted Apr 1, 2010 6:52 UTC (Thu) by mitchskin (subscriber, #32405) [Link]

I was speaking a little loosely; by "encapsulation" I was also referring to things like process isolation. So it's also like programming a computer without an MMU.

Open-source biotechnology

Posted Apr 1, 2010 14:40 UTC (Thu) by nix (subscriber, #2304) [Link] (2 responses)

Quite so. Note that there *are* modular components in biology: things like
the Hox genes are conserved because so much other stuff depends on them
that if they change, the organism dies. But conserved components like this
are often little more than genetic switches, at most generating protein
that binds to regulatory regions of the DNA: the things that actually do
the *work* are rarely so conserved. (Sometimes they are, but even truly
ancient and insanely well-conserved things like hsp83 occasionally
mutate.)

More generally: conserved stuff has lots of stuff depending on it, so is
often hard to use in isolation; non-conserved stuff often depends upon
conserved stuff, so is often hard to use in isolation.

Open-source biotechnology

Posted Apr 1, 2010 16:45 UTC (Thu) by flewellyn (subscriber, #5047) [Link] (1 responses)

I understand that programs created by genetic algorithms often have similar issues, as far as interdependency and nontrivial interactions between components go. Do you know if this is true?

Open-source biotechnology

Posted Apr 2, 2010 13:47 UTC (Fri) by nix (subscriber, #2304) [Link]

Yes, and for similar reasons. But GA programs have tiny population sizes
and tiny runtimes compared to the sizes and runtimes which have brought us
present-day biology, so they have had less time to accumulate arcana.

Open-source biotechnology

Posted Apr 1, 2010 16:44 UTC (Thu) by flewellyn (subscriber, #5047) [Link] (2 responses)

Yeah, I get all that. I just thought nix had made a mistake attributing the protein folding problem to DNA.

But, it seems I was wrong on that score.

Open-source biotechnology

Posted Apr 2, 2010 16:57 UTC (Fri) by nix (subscriber, #2304) [Link] (1 responses)

Well, technically they're disconnected, but if you want to make DNA that
*does* anything useful, you need to solve the protein folding problem as
well, because the machines that DNA builds are all protein-based. (I think
the only stuff this doesn't apply to is siRNA, which uses antisense
matching, which is of course trivial, particularly given how short siRNAs
are.)

... actually, that's an oversimplification. Some machinery DNA builds
(generally really ancient stuff) is catalytic-RNA-based (e.g. parts of the
ribosome). However, *that* gets its function in the same way as protein
does: it folds into strange 3D shapes, and those shapes have their effects
via the strange intermingling of chemistry and quantum physics which
affects interactions at that scale. So that doesn't help you much. You
still have the horrible folding problem to solve.

Open-source biotechnology

Posted Apr 2, 2010 19:58 UTC (Fri) by mjg59 (subscriber, #23239) [Link]

The last few years have shown that the role of miRNAs are more significant than people had
previously expected, so I wouldn't go so far as to say that the problem domain is pretty much
limited to that impacted by the protein folding problem - on the other hand, I think this is pretty
way off topic at this point...

Open-source biotechnology

Posted Apr 1, 2010 11:19 UTC (Thu) by cowsandmilk (guest, #55475) [Link] (3 responses)

I believe in Jim Collins's talk Biology by Design
(http://www.bu.edu/phpbin/buniverse/videos/view/?id=261), he describes the differences in the
unrealistic expectations many people had for biological toggle switches and what he created. The
time delay was just enormous and it continues to be due to the nature of biology. Nucleotides may
"encode two bits", but they most definitely can't contain the information of two bits. And there's no
way without inventing whole new ribosomes or completely novel sigma factors to change this. No
lzma compression to the genome.

Open-source biotechnology

Posted Apr 2, 2010 13:45 UTC (Fri) by nix (subscriber, #2304) [Link] (2 responses)

Unfortunately that's a Flash stream with a name disguised in a mass of
JavaScript. I have no idea how to play one of those without the Flash
player: maybe rtmpdump could do it, but even then I'd have to unpick the
URI. No thanks.

Open-source biotechnology

Posted Apr 7, 2010 0:05 UTC (Wed) by Baylink (guest, #755) [Link] (1 responses)

If you're on a Linux box, the Flash player will generally dump the entire FLV file in /tmp, and you can move it wherever and play it with a current mplayer whenever you like.

Open-source biotechnology

Posted Apr 8, 2010 16:01 UTC (Thu) by Duncan (guest, #6647) [Link]

Except that, on a site such as LWN where folks are more likely to be running
Linux boxes already, many people won't have your "magically assumed to be
there" Flash player at all, as it's proprietary, additional to the fact that
there's a lot of platforms without official flash binary support even if
people don't have issues with such things as granting damage waivers to
black-box code the authors seem to expect people to just trust that they
don't do anything harmful with, despite the fact that they don't respect the
same people's rights enough to open the code such that it can be fairly
examined before granting such a waiver, etc.

So the polite thing to do on a site such as this, is link the FLV file
directly, if possible, or if not, at least mention that the link is to a
flash video, and that if anyone wishes to extract the direct link to the flv
and post it (or conversely, extract the data itself and post that, assuming
legal permission to do so being granted, of course), it'd be appreciated.