Cheap DNA Sequencing Is Here – Writing DNA Is Next
wired.com
wired.com
The key enabling technology is the ability to (in-vitro) turn embryonic stem cells into gametes. This has been done in mice, but not humans.
The paper you link is pretty handwavy regarding that point; Bostrom seems tho think that if we just had more data we will be able to find a model that links genotypes and phenotypes. However, this is far from certain. There are too many degrees of freedom in our genome, and there are too many factors that contribute to something generic like "cognitive ability". Each individual contribution is so tiny that the effect is lost in noise, even if we sampled all humans.
And even if someone comes up with a novel statistical method and finds a reasonable model, optimising that model would be pretty dangerous. If you just optimise for one trait, the chances you'll even end up with a viable embryo after a dozen iterations are pretty slim.
There's a lot of selective pressure on cognitive abilities; so there must be reasons why we aren't smarter than we are. These reasons will probably kill your experimental embryo.
Maybe for some genes, but for thousands of others, the signal is definitely there. Intelligence is highly heritable –as much as height. Just like height, it's spread across thousands of genes. Yet GWASes have found hundreds (soon to be thousands) of genes for height.[1] If researchers had a quarter of a million pairs of IQs and genomes, they'd find a similar number of genes for cognitive abilities.
> If you just optimise[sic] for one trait, the chances you'll even end up with a viable embryo after a dozen iterations are pretty slim.
That seems highly unlikely. We've done far worse with domesticated animals and they've been quite viable. Some breeds of dog are predisposed to genetic diseases such as hip dysplasia, but that's because we didn't have the ability to screen for genetic defects. (Or more cynically: breeders cared more about looks than diseases.)
When domesticating animals, breeders often found rare traits and exaggerated them. But for human cognitive abilities, the best alleles are already prevalent in the population. It's just that nobody has lucked into all of them at once. And unlike height, there's no square-cube law to disadvantage more cognitive horsepower.
> There's a lot of selective pressure on cognitive abilities; so there must be reasons why we aren't smarter than we are. These reasons will probably kill your experimental embryo.
Humans have definitely been selected for intelligence, but we were also selected for other things such as famine resistance. The tradeoffs are different today, and we can do a much better job than nature. We can fix a lot of mutation load. Really, we're not nearly as smart as we could be. In the words of Nick Bostrom:
> Far from being the smartest possible biological species, we are probably better thought of as the stupidest possible biological species capable of starting a technological civilization—a niche we filled because we got there first, not because we are in any sense optimally adapted to it.
1. http://www.nature.com/ng/journal/v46/n11/full/ng.3097.html
Concerning your comparison with breeding: you seem to misunderstand me. I don't think that breeding per se doesn't work. Breeding by looking at phenotypes works wonderful.
But what you suggest is to just iterate to optimise [according to my spellchecker that is how you spell it] according to a model. You are missing the step where those embryos grow (or fail to grow) into adults. Without this step, your optimisation procedure will drift randomly on every axis that your model doesn't measure.
Are you sure? http://www.bbc.com/future/story/20150413-the-downsides-of-be...
I have a pet theory that while a modest intelligence advantage is likely to be genuinely beneficial, social animals like humans quickly run into trouble if they stray more than a couple of standard deviations from the average. Imagine living all your life surrounded by idiots (parents, teachers and superiors included), having to explain the stupidest thing in painstaking detail (and regularly being scorned as the one who doesn't get it), and of course having no palatable romantic partners. Down that road lies bitterness, voluntary solitude and maybe substance abuse. Something like http://nautil.us/issue/21/information/the-man-who-tried-to-r...
I think this is the most accurate description I've ever seen of my pre-university life.
Anders should aspire more towards George Church and less towards Nick Bostrom. (I know he wasn't on the paper, but whatever. It's been on my mind.)
While I am at it: here's some (typical disclaimer applies) notes on open-source DNA synthesis machine design: http://diyhpl.us/wiki/dna/synthesis/notes/
Re: Genotype/phenotype; once we find stable phenotypes, we should make more reliable structures on top of those phenotypes, instead of messing with the genetics. For example, for protein-based molecular nanotechnology, nobody really wants to spend their time doing a trillion different folds in simulation or in the lab just to find their target structure. Instead you will have to skip tweaking genotypes (because of the large costs of exploring the genetic landscape (really, take whatever you can get)) and just use lego brick phenotypes... at least for nano/molecular structures (nanotech). Other changes require exploration of evolutionary/genetic landscape, until heuristics get better if ever, which will continue to be expensive even with $1/genome synthesis costs.
Of course, knowing all protein (structural) phenotypes for all amino acid sequences up to length=100 would also be nice... but seems unlikely something we can work towards due to constraints of scarce universe.
My opinion: Bostrom has some interesting perspectives, but has always tended to preach pseudoscience to an audience that is entirely non-critical.
Not the same, but quite related: https://www.youtube.com/watch?v=2JNSpMhJLvg
Actually... quite possibly. Check out the first slide in [1]. You can basically build genetic logic gates using promoters to conditionally (i.e. depending on the presence of light, a protein, etc.) produce a protein to induce or repress another promoter, thereby increasing or decreasing the expression of another gene.
It's still pretty early stages and there are a lot of factors that make this more complicated (life isn't binary), but the ultimate goal of synthetic biology is to be able to program DNA in the same way you might program a computer.
[1] http://co.mbine.org/events/COMBINE_2015/agenda?q=system/file...
DNA (with supporting cell machinery) create us, so...
Identical twins have the same DNA, but they don't have the same fingerprints. Why? Different environment in the womb.
Could we tweak things like eye color? Sure. Things like intelligence? Much, much harder.
TL;DR: read "Reflections on Trusting Trust", the same phenomena applies to biology.
The gap in understanding between 'here are a bunch of bits that appear to do 'x'' and 'x' is vast, far larger than the gap between say being able to build a compiler to have an executable perform in a certain way.
If anywhere that is where the real breakthrough will be, in being able to effectively map the chain of binary bits in a genome to a certain phenotype in a mature organism without any 'leftover bits'.
Once you have that chain of causality mapped out in enough detail to clearly link all causes with their effects you can start engineering something to act in a way that is far outside what we normally find in nature.
I'm not sure if I'm expressing myself clearly about all this, maybe someone more versed in the matter can chime in but the path between DNA and organism is a lot longer and less well understood than the path between a low level language program and the resulting output of that program. Just having a 'compiler' will not change much about that understanding, though once we have that compiler we may be able to advance our understanding by a more direct approach in modifying pieces of an existing genome to observe the effects in the organism. For sure it will speed things up.
edit: if we did have that understanding it would be trivial to create a digital representation of a (simple) organism from it's sequenced DNA. The fact that we are nowhere near capable of doing this is a simple proof that merely being able to create a certain sequence of bits will not give us the capability to create organisms to order.
Bacteria are "simple" enough that we can just append some DNA sequences and have them develop stuff we want. It's predictable enough that people already created "standard library" of DNA blocks - so people can work with bacteria in a kind of dataflow environment, with underlying DNA representation hidden. You connect a "molecule sensor" block with an "amplifier" block, an "inverter" block and a "light up" block, and you get bacteria that start to glow in presence of some metal, etc.
But of course, once we move to the level of multicellular organisms, the complexity explodes. Here, as you say, we basically know that something in an unaltered system sorta correlates with a particular phenotype and if we change it then maybe it won't blow up.
The compiler is called "genetic engineering" and it's a 30 years old field. In particular genetic engineering allows you to knockout specific genes and observe the phenotype. This technique is commonly used on bacteria, plants and animals. There is an ongoing mouse knockout project, an attempt to study every mouse gene by turning it off. It is 50% complete so far. By the way, the experiments are performed mostly by hand.
>We're very far away from understanding how a particular DNA string relates to phenotypes.
This problem can be solved much faster with a large scale effort. Thousands of automated experiments can run in parallel while recording the phenotype of the animals (including results of behavioral tests) and storing it for later analysis. On this genotype->phenotype dataset one can train an ML model. Actually there is a company that is trying this approach (just ML part, without automated experiments) http://www.deepgenomics.com/
I doubt that one company will be enough, though. With serious funding there could be much more progress in this area. We could have a full knockout map of mouse genome in less than a decade if we really wanted.
Even for a small bacterial genome (1Mbp) that's $10,000 at a penny a base. To have any real power, I'd want to build at least thousands of them which is a little out of my price range.
2) DNA synthesis has been tracking Moore's law for a number of years, and doesn't show the signs of slowing down anytime soon unlike transistors.
EDIT: I know what's a base pair. I didn't understand the "penny" and "dime" in this context.
As others pointed out, writing DNA efficienty is the bottleneck. Reading is getting cheap.
That's a bit hyperbolic, isn't it? We understand the basics many organic systems and our knowledge and tool sets are improving at an accelerating rate. Look at the recent gene therapy used to cure leukemia: http://www.techtimes.com/articles/104545/20151109/babys-leuk...
To my eyes writing DNA will only be a spectacular show of nopes
But we'll be able to learn so much from the failures. We'll finally have the coding and debugging tools to make a serious stab at reverse engineering life.
The sooner we get real insights and are able to find new cure the better. We'll see.
We engineer machine intelligence just enough for it to spark (tipping point, runaway, whatever you want to call it), and then it'll engineer us "better". Is "better" what we consider better? Or what it considers better? Interesting times.
Or, rather, how does copyright work for genetic designs? Syn Bio papers are published all the time, and their designs are copyright like any other invention.
(pun not intended but given the political circumstances the expression does give pause for thought, maybe a better term would be:)
Determinism over randomness.