I’ve been working on a synthetic DNA assembly company. Basically, I figured out how to assemble DNA for people at a fraction of what it normally costs, so they give me a sequence, and then I make it in real life for them, then ship it to them.
Most of my customers have been AI protein designers, ironically. Turns out SOMEBODY has to wrangle atoms in the real biological world and that’s me!
After almost a year of work I finally smoothed out all the kinks in the process, so can now go from a design to synthetic DNA in a cell in about a week (not counting oligo pool synthesis time). I can do about 600,000bp per week, which is large enough to synthesize the smallest bacterial genome (each week), tho I only do about 1000bp fragments. I’m also completely bootstrapped and self funded, and only get help from my several opentrons robots
I once met a freelance bespoke industrial adhesive maker. He takes orders from various factories for adhesives with specific properties, then uses his knowledge of chemistry + trial & error to make one that fits the specs provided.
A bespoke freelance industrial adhesive maker sounds like such a niche job, that's awesome. I would love to see a hackernews-type post with details of how he thinks about making a specific adhesive.
I can share a few things! (but definitely not all)
- One is finding different T7 RNA polymerases with unique properties by manipulating the backbone. They can be used for things like in-vitro RNA production for vaccines
- Another is synthesizing a phage that has been sequenced for a specific organism, but that the samples are now lost of. So resynthesizing that genome from scratch
- A different project (personal one) is building a DNA parts toolkit with standardized DNA parts so you can combine em together like legos. Pretty much nowhere but FreeGenes has open source genetic parts (I used to run that project), and I think open source genetic parts need to be in the world
I sometimes derive protocols from off-shelf ones, but pretty much everything beyond that is in-house. Most off-shelf protocols work for 1 sample in 1 tube - I had to adapt them to working on 1 plate of 384 tubes (and get those to work with robots). There is a significant amount of robotic code that I use, and a few custom protocols that are from a random obscure scientific paper in 1980s or 1990s
Synthetic DNA for data storage seems much higher to degradation (ie, heat, light, or other sources of radiation). Not sure if I would use syn DNA to for anything long term.
The oldest sequenced DNA is 1.6 million years old. In the right conditions, it lasts far longer than pretty much any storage method right now. Plus I also sometimes store things in Bacillus subtilis, which is very hardy https://keonigandall.com/posts/sporenet.html
You are asking for a counterexample
but I think s/he was asking for what assurances the software can provide. Is it proven to be perfect, bug free in both specification and the physical analysis?
Remember compared to you assume we
are all laymen.
It would be interesting for us on the armchairs how such software even works. Is it like (ahem!) antivirus software looking for patterns?
The reasons I ask for a counter example is because the number of “what ifs” are pretty much unlimited and easy to come up with, whereas I have to do a lot more thinking to answer them. It’s asymmetric and gets tiring
Answer is: no, not proven to be perfect. Most source is closed for the software. There is no specification. Never been a red team trying to break it. Unknown if it’s ever stopped any threats. There is no requirement to do it, and it’s all voluntary (though pretty much all synthesis providers do it). Mostly they just hash kmers and translations thereof and scan a pathogens/functional database. So implementable in a simple KV store for the most part. The bigger problem is false positives, and most work is done in de-shittifying the upstream pathogens data.
Last I checked all the groups charged a lot of money for access, unfortunately, and I’m not really big enough to join a group like the International Gene Synthesis Consortium :(
That's how Twist/Agilent do oligo pool synthesis (and Dynegene now). I'm pretty interested in the Genscript / Avery digital method of electrochemical synthesis. Turns out those pools ain't good enough to be used in a biological context, which is where I come in - I can assemble them well into sequence perfect stuff
I found a virology textbook at the local Catholic book fair when I was in 6th grade, got hooked, teacher in 7th grade let me order GFP transformation kits to the school that I could do at home, then off to the races from there. I was in this article if you're interested in more deets: https://www.nytimes.com/2018/05/14/science/biohackers-gene-e...
That's awesome. I also got hit hard by the science bug (pun intended) as a kid, which I can partly attribute to finding a random virology textbook in an academic bookstore dollar bin. I was obsessed with virology at least until college. I ended up majoring in math in college but made it through Ochem II and did some lab internships before committing to that path. Now I'm a ML/software engineer with a healthy interest in biochem.
Cool! Though it's a bit scary how easy it would be for someone to make something dangerous as mentioned in the nyt article. The 2018 article ends:
>“There are really only two things that could wipe 30 million people off of the planet: a nuclear weapon, or a biological one,” ....
>“Somehow, the U.S. government fears and prepares for the former, but not remotely for the latter. It baffles me.”
I guess we may have seen that kind of thing happen a year later with covid though that would have been government sponsored mucking around rather than DIY if it wasn't natural. Not sure how we stop that happening again?
Personally, I think the threat from the biologics itself is a little overstated. If COVID was released from a lab (I think it was natural), it was most likely due to bad governance and management of the lab. For DIY things - most DNA can be screened. I have had companies offer screening services for about 50k-100k a year. I can't afford that! So hopefully a free service or something near to that comes online.
I never went to university or anything, but did work for 4 years at UCI in directed evolution / mitochondrial engineering, then 3 years at Stanford on the FreeGenes project (got invited to work at both). Barely passed high school cause I didn't care about my classes there.
The key is doing that at an industrial scale, reproducibly, with hundreds to thousands of plasmids at once. Becomes less simple. You encounter bullshit problems with biology, which I guess is valuable because it makes a moat!
Most of my customers have been AI protein designers, ironically. Turns out SOMEBODY has to wrangle atoms in the real biological world and that’s me!
After almost a year of work I finally smoothed out all the kinks in the process, so can now go from a design to synthetic DNA in a cell in about a week (not counting oligo pool synthesis time). I can do about 600,000bp per week, which is large enough to synthesize the smallest bacterial genome (each week), tho I only do about 1000bp fragments. I’m also completely bootstrapped and self funded, and only get help from my several opentrons robots