De Novo DNA synthesis by Mondoplex PCR

A big challenge in the discipline of synthetic biology is the fast, error-free, and inexpensive creation of long strands of DNA from scratch. Traditionally, the enzymatic modification of the DNA sequence by PCR required the presence of a template DNA sequence that was in large part similar to the desired sequence, for only a few nucleotides could be modified at once.

Over the last two decades, life science companies have been offering services to synthesize custom strands of DNA. These "gene synthesis" methods generally work through the assembly of many small DNA fragments (<100 bases each) into larger fragments of up to several kilobases. Although the technology has improved steadily, in many cases it is still prohibitively costly to synthesize DNA de novo. (I believe the best bang for the buck is around 15 American cents per base, depending on where and when you place your order). As I understand, the phosphoramidate chemistry that is used to assemble the synthetic DNA strands has an error rate of about 1 in 1000 at best, meaning that synthesis of strands larger than a few hundred bases becomes rather improbable. In contrast, DNA polymerase enzymes that copy DNA from a template can have error rates as low as 1 in 10 million.

A platform that offers one the freedom to synthesize any DNA sequence with the fidelity of enzymatic DNA polymerization would be most desirable, and the commercial license of such a platform would be most lucrative. Most companies will synthesize your DNA on demand when you order it. By harnessing digital microfluidic platforms that allow one to "move" droplets of biomolecular solutions around, however, the throughput of the reaction can be massively increased if we synthesize ALL of the 15 base pair sequences at once, and then copy from the ones that are required for assembly of the desired sequence. Since the service is no longer requires DNA synthesis on-demand, rather DNA copying on-demand, all DNA polymerization can be carried out enzymatically, drastically lowering the error rate. The device whose construction I am proposing is based on established molecular and synthetic biology methods that have all, to my knowledge, been commercialized in one form or another.

The process is based on iterative DNA polymerization cycles between a growing primer sequence of indefinite length that occupies its own droplet, and millions of different 15bp template sequences, each of which is tethered to its own reaction chamber in multiple copies. These template sequences, having been synthesized chemically and purified by liquid chromatography are error free, and their 3'-most nucleotide contains no 3'-hydroxyl (or are otherwise chemically modified) so that they themselves do not act as primers and extend themselves. The digital microfluidics platform guides the droplet containing the primer sequence to the next reaction chamber, where the enzymatic synthesis reaction occurs. When the primer sequence is of the desired length, it is brought to the final reaction chamber where it is amplified to a suitable concentration by PCR.

Given that there are 1 billion sequences of 15 base pairs in length (15^4), and assuming that each reaction chamber is one cubic centimeter (probably a huge over estimate), then this DNA synthesizing monolith would be a cube with a 10 m side length. Of course, given the similarity of each DNA sequence to some other DNA sequence, you could probably throw away about ((4^2)-1) / (4^2) of the sequences (~90%) without increasing the travel distance too much. If a droplet has to travel on average 1 meter to the next reaction at 1 meter per minute, and the reaction takes 1 minute (probably an overestimate), then about 10 bp can be assembled in 2 minutes given a 5 nucleotide overlap sequence required to prime against the template. This equates to around 1 kb per day, which is modest but because of automation, many droplets passing through the machine at once, the economies of scale blah blah blah you get the idea... My gel's finished, I have to go.