[corwinbad]

Using Twist Bioscience DNA - interesting data points: 215Pb/Gram ~1,300 copies per unique oligo enough to get a 100% data back Almost unlimited coping of recoverable data using PCR

Now only need to increase the throughput and price of DNA synthesis by 4-6 orders of magnitude

[jlj]

What possibilities are opened up when 215Pb/Gram with good I/O and price is reality?

With computing power increasing 1 order of magnitude around every 5 years (Moore's law), this might be reality by 2027.

An antecedental story is computer graphics, it only took 20 years for to go from idea to reality. Learned a lot listening to a talk at work by Alvy Ray Smith a couple years back. The true poineers are building now not for what is possible now but for what will be possible tomorrow. This article gives some context https://www.forbes.com/asap/2001/0528/052.html

This is another interesting perspective on time scale of orders of magnitude of computing power https://en.m.wikipedia.org/wiki/Computer_performance_by_orde...

[gwern]

> What possibilities are opened up when 215Pb/Gram with good I/O and price is reality?

The data storage is almost totally irrelevant. When you improve DNA synthesis by 3-4 orders of magnitude, you make it dirt-cheap to synthesize whole custom genomes, hyper-optimized for anything you wish. (Currently, the state of the art is synthesizing an E coli genome takes multiple years and a large international collaboration.) CRISPR only lets you do a few edits at a time at most, in some parts of the genome; imagine being able to do tens of thousands of edits as easily as 1 edit. At that point, you're hardly even 'editing', you're designing organisms as a whole. The Church lab has some astounding proposals for things you can do when you are able to synthesize a whole genome, like cleaning out all the retrovirals, movable elements, recoding the genome for total viral immunity, on top of the obvious stuff like eliminating all mutation load or increasing specific traits by dozens or hundreds of standard deviations. The mind boggles.

[jgill4]

Craig Venter, Clyde Hutchison and Daniel Gibson produced a minimal genome in a relatively short period of time and even transplanted it. This was a "bottom up" approach in which all of the DNA was synthetic. This is a trivial thing for this team now. As you mentioned, when synthesis costs drop dramatically, things will get interesting!

[frozenport]

Another viewpoint is that in an academic setting, there is no reality-check, professors rarely loose their jobs, students are free, and operating budgets are small enough to be overlooked. The only people judging are your peers, who often have the same kind of nonsense scheme. Even if it doesn't work, and is completely useless, the idea of DNA storage is guaranteed to never die.

[tiatia]

"Almost unlimited coping of recoverable data using PCR"

Actually, this would not be as trivial as it sounds. Especially the unlimited coping.

[jgill4]

It's actually not hard to do. In theory, a single PCR reaction with only two primers could copy the entire data set in a few hours.

[tiatia]

Are you sure you understand the concept AND the limitations of PCR? I predict: No.

There are many forms of DNA. Single stranded, double stranded, circular etc. etc. I assume we were talking about non-circular DNA. ds/ss does not matter, for PCR you would have to go ds anyway. You realize that strand shortening would be a problem? I could think about ways to correct this but it is not trivial.

[blurbleblurble]

Yeah, but just wait until some clever bacteria figures out how to hack your code for selfish purposes!

[dnautics]

what's the error rate on copying DNA using PCR?

what's the fastest you can read out DNA using sequencing?