DNA synthesis is becoming a bottleneck for screening AI protein designs, designing CRISPR-Cas libraries, and high-throughput cloning.
Can biology and electronics join hands to solve this problem? Today’s paper takes a stab at the problem!
Don’t keep this newsletter a secret: Forward it to a friend today!
Was this email forwarded to you? Subscribe here!
DNA Goes Parallel
Researchers created a new technology to synthesize DNA by combining electronics and biology.
Learning to Write DNA
DNA encodes the information for all life.
And thanks to sequencing, we now understand a huge amount of it. From Sanger sequencing to next-generation platforms like Illumina or Nanopore, we might be in the golden age of DNA reading!
But DNA synthesis?
Sometimes, it feels like the Paleolithic. Orders might take weeks, some sequences are just too hard to make, and it becomes very expensive, very quickly. Any scientists I know are masters at balancing price, speed, and sequence complexity!
To be fair, we’re great at synthesizing short, simple sequences (< 150- 200 nt).
And this powers most of modern biology: the primers in your PCR, the cloning used to create insulin and GLP-1s, and high-throughput diagnostics. DNA synthesis touches every aspect of biotech!
But if synthesis were easier, faster, and more parallel, we could do much more!
Where DNA Synthesis Gets Stuck
Okay, but where’s the bottleneck? Well, there are two main ways to create DNA from scratch without a template:
Chemical synthesis
This is the classic option. IDT, Sigma-Aldrich, Eurofins: they all use this. It relies on phosphoramidite chemistry, where individual bases are added to a growing chain on a solid support. It’s precise and scales well! The drawbacks? Efficacy and accuracy drop >200 nt, and the massive use of dangerous solvents.Enzymatic synthesis
On the other side of the spectrum, we have enzymatic synthesis, mostly offered by smaller companies. It’s usually based on terminal deoxynucleotidyl transferase (TdT) to add one nucleotide at a time and uses mild, water-based reagents.
So, in short:
Chemical synthesis scales well, but uses horrible reagents.
Enzymatic synthesis uses mild reagents, but it’s harder to parallelize.
Why does parallelization matter, you ask? Many applications need it. I already mentioned clinical diagnostics, but there are also guide RNA screenings for CRISPR-Cas systems, library screening for AI-based protein designs, or DNA data storage.
Everything starts with DNA!
Parallel, Enzymatic DNA synthesis
Today’s paper asks: Can we parallelize enzymatic synthesis?
And well, they did it! The team combined electronics and enzymatic synthesis to synthesize 64 DNA sequences at the same time. That’s a 5x improvement over previous results (at least!). And with no toxic solvents!
The core idea is to use a complementary metal-oxide-semiconductor (CMOS) chip as a programmable synthesis surface. The chip controls localized chemistry at many tiny sites, letting the authors deprotect DNA in selected places and extend it enzymatically, one base at a time.
How does it work? Let’s take a deeper look!
CMOS + Enzymes: Localized DNA Synthesis
This is cool.
The heart of the system is a CMOS integrated circuit. On top of it is an electrochemical array formed by 256 (16×16) “pixels”. Each pixel is a concentric pair of anode and cathode ring electrodes, with a gold center where DNA is immobilized. This DNA is protected from elongation by chemical modification.
The CMOS is programmed to send current between the cathode and anode, lowering the pH around the immobilized DNA. This acidity indirectly removes the chemical modification, and now TdT polymerase can extend the strand!
This is the basic cycle:
Fabricate the 256-pixel CMOS array with normal electronics production.
Deposit gold in the center of each of the pixel.
DNA is immobilized using gold-thiol chemistry, with the other end protected by a chemical modification.
An electric current lowers the pH to ~5.3 and deprotects the DNA.
Extend the strands enzymatically, adding a new protected nucleotide.
Repeat until you reach the target length.
And voilà! An elegant solution to the problem. And to make it even cooler, microfluidics handles the addition of buffers and reagents. No more pipetting errors!
On To The Synthesis: One to 64
Single-sequence synthesis
The team first tested a simple but important benchmark: a single sequence on all 256 pixels. The authors built a sequence with a 10nt feature sequence, plus a universal 23nt primer (for downstream amplification).
Then, they compared two ways of doing the deprotection step:
electrochemical pH localization ← their method
versus bulk acidification at a uniform pH ~5.3
For this design:
Both methods synthesize the 10nt feature sequence with 0 errors!
The electrochemical method is noisier, with the bulk acidification giving more reads and a sharper distribution.
But it worked! Great. So, they moved to longer feature sequences:
20 nt → total length 43 nt
59 nt → total length 82 nt
79 nt → total length 102 nt
As the length increased, they saw
The length distribution broadens
The per-position errors increase
But the sequence consensus still matches with no errors!
But the story is different when we look at the full-length (not only the feature sequences). The full-length purity goes from 52.8% at 10 nt to 2.2% at 79 nt, and the per-nucleotide error rises from 9.6% to 42.18%! Interestingly, the errors are mostly deletions.
Multisequence synthesis
But the real demo is the parallel synthesis of 64 different sequences!
To reduce crosstalk, the authors used every other pixel. From 256-pixel array → 64-pixel synthesis. Each pixel can be programmed to activate at different times, controlling the synthesis.
Each target sequence contains:
A unique 5nt ID → lets them map each sequence to the pixel that synthesized it.
A unique 10/11nt feature sequence.
A common 23nt primer.
The result?
The synthesis of the 64 distinct sequences worked
For the feature sequence, there are 13 wrong bases out of a total of 676 synthesized bases
On average, the per-nucleotide error is around 30%
Now, are these the best synthesis results ever? No. far from it! But they’re pretty decent for the first generation of a new parallel enzymatic synthesis system!
Demonstrating Data Storage
The 64 sequences produced in parallel aren’t random ones.
They encode a 169-byte quote by Einstein! The recovered text is practically entirely correct, with just a small error (“entire” → “entRu”). This is a pretty strong proof-of-concept for parallel DNA data storage! Not the most mind-blowing, but it’s pretty cool that even a first-generation system like this can encode data with minimal errors.
Advancing DNA Synthesis
A pretty cool paper!
I used a lot of synthetic DNA in my years in the lab. And if you’re doing non-standard research, there are many bottlenecks. Long or complicated sequences are challenging and expensive!
Is this system going to change it tomorrow? No. But it’s amazing to see DNA synthesis in the spotlight again! I felt like it was all sequencing for a long time, and no one cared about those poor scientists cloning on the weekends.
And the authors here show that the limitation of their method isn’t the hardware but the chemistry. Their pH-based deprotection systems create intermediates that damage the DNA at every synthesis cycle.
Improve the chemistry, improve the system! I’m sure they’re working on that already. And using CMOS is another advantage: it’s a mature technology; we know a lot about it, and we know how to scale it! Not like biology…
And better DNA synthesis could enable better clinical diagnostics, more exciting cloning, and new DNA chemistries! Plus, enzymatic synthesis is better for the environment: a win-win!
So, go here, read the paper, and get inspired!
If you made it this far, thank you! What do you think of combining electronics and biology? Do you think it has a place in biology? Reply and let me know!
P.S: Know someone interested in DNA synthesis and SynBio? Share it with them!
What did you think of today's newsletter?
More Room:
Loading DNA Origami: It’s common to functionalize DNA origami. How does that impact the structure? In this paper, the authors show that cationic peptides bind non-specifically to DNA origami, leading to unpredictable loading and reduced functionality. They demonstrate that extensive PEG purification and site-specific peptide attachment are essential for achieving precise functionalization and reliable therapeutic performance of DNA nanodevices.
Cooling Down RNA: Yes, we’re talking about nucleic acid preservation again! This review examines strategies to improve the thermostability of mRNA-LNP vaccines, reducing the need for ultra-cold storage. It discusses approaches including mRNA engineering, lipid nanoparticle design, formulation optimization, and alternative stabilization methods such as lyophilization and deep eutectic solvents, highlighting that combining multiple strategies will likely be necessary for stable room-temperature vaccines.
From (DNA) Origami to (DNA) Kirigami: Kirigami is a variation of origami that adds cutting to folding. In this paper, the authors develop a kirigami-inspired design strategy to program the self-assembly of DNA origami into curved crystalline structures with diverse shapes, including toroids and helical tubules. This approach enables precise control over crystal geometry and provides a general framework for engineering complex 3D nanomaterials.