Re: Genetics, nannotechnology, and programming

Robert J. Bradbury (
Fri, 22 Oct 1999 20:49:29 -0700 (PDT)

On Fri, 22 Oct 1999, Skye Howard wrote:

> what about engineering some of those artificial chromosomes
> to create artificial structures- for example, an implant
> of some kind that would be inherited.

Unless you put the AC into the germ line cells it will not be inherited. However doing this is what Greg Stock and John Campbell are proposing. (By puting an AC into the first cell of an embryo, it will be inherited by the germline cells as the fetus develops.

As far as ACs creating hard nanotech structures it is doable. Think, teeth & bones & sea shells. However since these are patterned on the scale of eukaryotic cells (10 micron scale) and the smallest we could probably go is bacteria (1 micron scale), it is doubtful you could get a cell to make a nanoassembler without a *lot* of work designing new genes that would self-assemble into an assembler.

> I mean, for example, you could enter the
> specifications into a computer as the "desired output"
> and then running it through one of those genetic
> algorhythm programs. What better application for
> genetic algorhythm programming than genetic
> engineering?

This is similar to what I have proposed in my NanoParts@Home scheme. If you have a description for something and a SETI@Home type distributed program that can generate random collections of atoms and do genetic evolution of the designs until they meet the criteria specified by the description.

So, I could use this to "evolve" an "atomic" rivet or an enzyme with a known structure-function. I can't use it to evolve a nanoassembler because we don't have a complete description for a nanoassembler. (So the genetic evolution algorithm has no way of selecting the design "successes").

> A) Genetic engineering of mechanical devices (possible
> jumps on nannotech using cell replication methods?)
> (inheritable implants?)

Almost everything going on inside cells involves "mechanical" devices at the molecular scale. The reason enzymes work is that they position molecules in close mechanical proximity and apply directed energy (chemical or mechanical) to drive reactions forward much faster than they would normally occur.

It is best to keep separate 5 things:

(1) molecular nanotechnology (atomic or small molecule manipulation)
(2) self-assembly (molecules that put themselves together)
(3) self-replication
(4) easily programmable nanoscale machines
(5) wet/soft vs. dry/hard nanomaterials
(6) self-modification

Cells do 1-3 with wet nanotech. What nanotechnology is all about is doing 1-4 with dry/hard nanomaterials meaning you have to solve 1 & 2 with an entirely new toolset and assembly process. We are trying to compress what nature took a few billion years or so develop into a couple of decades.

It worth noting that *most* of the economic benefits attributed to nanotechnology can be obtained by mastering (3) and perhaps (2). The degree to which (4) is required for this is open to discussion. Having (1) with hard nanotech materials is only the frosting on the cake. Interestingly enough, I think it is (4) combined with (6) on top of (1-3) in hard nanotech that raise all of the really nasty outcomes frequently discussed on this list. What you are suggesting in genetic evolution of hard nanotech comes dangerously close to that. It depends entirely on whether you have "safeties" in the selection criteria that only select variants that are not dangerous to their operating environment.

> b)applications of genetic algorhythm programming to
> genetic engineering.

This is already done without the "programming". There are routine experiments that are done in biotech labs where they "evolve" better enzymes or tools for specific purposes. It simply involves methods to create millions or billions of variants and then selecting the best of the bunch, mutating those and repeating. Generally it is called "directed evolution" or something similar and you can probably find dozens of articles on it in Medline.

Since in biology working with "billions" is easy while in computers it is still pretty difficult, biology will be a better approach than computers for a few more years.