Some ranting notes on schemes for microfluidic DNA synthesizers

Skip to first unread message

Bryan Bishop

Apr 24, 2009, 4:11:18 PM4/24/09
As the title says, here are some ranting notes on schemes for DNA
synthesizers in microfluidic devices. Thoughts would be appreciated,
although I'm sure I'll eventually think of something. There are a few
unknowns still (like why you'd want to use channels in an EWOD
setting), so nothing certain is coming from this.

There are multiple schemes that I have been thinking of for a digital
microfluidic DNA / oligonucleotide synthesizer. The first two went
something like this. (1) Circular DNA synthesizer: droplets go in a
circle and nucleotides and other chemicals are added to each of the
droplets. The problem with this method is that the volume of the
droplets increases per each merge step with any other chemical
droplet, so the droplets have to be very, very small-- two micron
diameter droplets are pushing it, because even then you need quite a
lot of area, or something. (2) Single nucleotide synthesis method. In
this method, you have four main channels. In each channel you
synthesize droplets when that nucleotide is present in that position
in the DNA molecule. So, if there is a droplet in the "A" channel for
position 930, then there either should be (A) no droplet in the other
three channels for that position, or (2) a droplet of just water in
the other three channels/lanes. Then, later, the droplets are
integrated such that position X = droplet from channel A + droplet (or
nothing?) from channel T + .. etc. And then you wrap each of those
droplets in the right chemical reactions to make it legitimate
phosphoramidite DNA synthesis.

(1) Circular DNA synthesis. Problem: volume of droplets increases on
each pass. Fixes the "layout algorithm" problem though :-). For small
genomes, a small droplet size will make it work. Somehow the
nucleotide droplets and so on have to be smaller than the 'synthesis
droplets'. Experiments are contained in a single droplet (neat!). The
volume problem can be solved by synthesizing only 20-nt oligos and
then storing them in a buffer. So, the droplets only increase to a
certain size on that pass; then, they increase to a larger size on a
secondary pass when all of the 20-nts are connected together. But the
volume problem stays the same. If you add the same amount of volume
for a 20-nt, and just split up the genome into sections of 20-nt
regions, then you're still going to get the same final volume. So the
only way to solve this is to just store large droplets of 'nearly
completed but very ridiculously large oligos' in giant resevoires
where you can afford to keep on adding in significant volumes of

Another solution for the volume problem in a circular droplet DNA
synthesizer is to just have an 'export region' where there are
"finished" DNA droplets. The finished DNA droplets are the ones that
are allowed to grow to very large sizes in a giant resevoire (but with
mixing or something). These droplets should grow from 1 micron to
however-much volume there is available. Do the calculations to see
whether or not it's a good idea. What this means is that instead of
moving around super-large-droplets, you're still moving around
relatively small droplets, and then just combining them with the large
droplets at a later point. At some point the droplets become too large
and they need to be kept in a giant resevoire where they will be
picked up later, in other words.

Do *not* move around the genomic DNA droplets. Those are supposed to
be getting large. Don't bother moving them around. Move the smaller
stuff around. Keep as small as possible droplets for each nucleotide.
So, if an entire "cycle" of DNA synthesis takes 200 picoliters of
volume, then

(2) Single microchannel-nucleotide DNA synthesis. Problem: ? Not very
highly parallel, I guess. You'll need to turn it into a four-channel
spiral structure to conserve space with the droplets or something,
otherwise you're going to need hundreds of meters for a 500 kb oligo

(3) Merge-based DNA synthesis. This is where you merge chemical
reactions ready to go each time. The merge channel mixes the
previously constructed oligonucleotide with the new base or new
chemical compound or something. Problem: layout algorithm. The
channels can't cross, yet they have to if you're going to try to get
everything to everything else.

(4) Split DNA synthesis method. Synthesize only segments of the DNA
molecule. Store these segments somewhere for later, and later combine
them with other strands. So, for a genome of size N, it should be
possible to split it up into a certain number of pieces such that the
individual pieces are small enough to be synthesized. Where do you
store the pieces? How do you keep track of pieces?

(5) Primers for extracting portions of an organism's genome that you
have found. i.e., no de novo synthesis other than for primers and
reacting with some genomic DNA. Since the likelihood of you
implementing a truly novel gene is going to be pretty slim-- although
the likelihood that you're going to have the genomic DNA laying around
that happens to have the protein or gene you want- well, that's also
fairly slim. So maybe this needs to be rethinked? Distributing an
entire biobrick library on a lab-on-a-chip ready to go?

(6) Completely wall-less synthesis. This would be controlled by some
sort of 2D grid for EWOD. The problem of DNA synthesis then becomes a
droplet routing algorithm problem. And if you go this route, then you
get to use those dynamic routing algorithms for other chemical
reactions in the future. Are channel walls necessary for EWOD?

It would be nice to have some equations to calculate maximum droplet
size per some size of EWOD electrode geometry, and then also how many
droplets you could do or what size droplets would be possible so that
we can evaluate which scheme is more possible or easier to do given
different types of technological capabilities.

- Bryan
1 512 203 0507

Reply all
Reply to author
0 new messages