Soylent raises $750k on Crowdhoster – what’s next for open source crowdfunding?

Basically the workflow is 1) find sequences that you want to stitch together. 2) build a draft plasmid. 3) make fine tuning changes (making sure RBSes are ok, we aren't stomping on terminators or promoters, etc, add mutations if we want). 4) annotate primer regions. 5) cut and paste into IDT's primer orderer.

The bulk of my gibson assemblies (about 50 of the ~100 that I've done personally, plus 12 that my college intern last year did, and 13 going on 48+ that my intern this summer is doing) are for site-directed mutagenesis; quikchange blows, and the plasmid I'm playing with is 20kb anyways, with 12kb of mission critical DNA. The general scheme is: design 2 mutagenic primers, 2 PCR using primers that flank a pair of unique restriction sites near our site, then gibson the left hand and right hand products with a stock of pre-digested plasmid. Since bacterial plasmid propagation is "error-free", we don't bother sequencing all but 1kb or so of the final product - saves on cost - and our plan is to do shotgun sequencing of all of the plasmids, barcoded, with ionTorrent when we're done.

We're also doing combinatorial installation of promoters that my boss' intern has done (27 constructs) - that I've partially supervised. Among the other things that I've done are gene fusions, leader sequence deletions, moving around restriction sites, swapping plasmid backbones, assembly of complex, highly repetitive sequences, and we attempted (but only partially succeeded at) capturing a eukaryotic chromosome by the telomeres.

We cut a lot of corners, but for NEB's sake (I like them) I'm not going to discuss exactly what those corners are, quite yet. Even with ours sloppiness, it works nearly every time, often with 8/8 correct (and sometimes you have to screen 48 to get one, but that's rare and I haven't had that happen in about a year or so). My general philosophy is the more sloppy you can get away with being, the more robust your procedure is - I like gibson assembly because I can teach it to an undergrad in a week and have them rip roaring on a project.

There are lesser known properties of gibson assembly (and this is in the original paper if you read it carefully, which most people don't) - you can have inexact ends, for example, if you want to have a template that is released using restriction enzymes - the 3' exonuclease activity of the polymerase will remove it and create a seamless construct with no evidence of the former restriction site. I have to warn: This would be hard (but not impossible) to implement in a design suite, and nearly impossible for automated design unless you are using exactly the same RE protocol each time.

Slight disclaimer - I'm a coworker of Dan Gibson's, so take my evangelism with a little bit of a grain of salt. I will also say, that there are plasmids that you absolutely can't gibson clone into and we don't know why. pBR322 tet marker is an unclonable position, for example.

Finally:, we don't really use computational biologists. Actually, we're designing enzyme variations from first principles and previous literature report, without appealing to simulation or model, with quite a bit of success.

I thought he was getting weekly blood tests?