Why Twist Bioscience’s complex genes offering is a bet on AI-driven protein design

Say you are in the market for form-fitted clothing for a special event. Off the rack won’t do, and your tailoring requests are demanding: structural alterations, unusual fabrics, tight deadline. Some tailors might balk at the request. Others might take your measurements, feed them into the system and tell you the job can’t be done. You might hunt down a specialist tailor who also sells high-end clothing. And end up buying your everyday wardrobe from the business as well.

That is roughly the strategic logic behind the latest product launch from the DNA synthesis firm Twist Bioscience, which is expanding its clonal genes line to cover longer and more structurally difficult sequences. The company says its new early-access offering, called Complex Genes, is expected to raise acceptance to about 99.5% of clonal gene sequences and 99.9% of all DNA products.

On Twist’s Q2 FY2026 earnings call, Patrick Finn, Ph.D., the company’s President and COO, put a number on its prior capabilities. “Three years ago, we accepted about 96% of clonal genes,” he said. At the time of the call, Finn said the firm accepted “about 97% of clonal genes.” He added that Twist could then manufacture approximately 98.5% of clonal genes and about 99% of DNA requests more broadly.”

The upper end of the market is still emerging. Finn framed the announcement as “going back to the true north of a really positive customer experience.” He added: “We see it as a very low single-digit percentage opportunity from a business standpoint in terms of market share. We shipped roughly a million genes last year, so it’s a nice bolt-on.”

The bet is that AI will make that bucket bigger. Generative protein design tools can now produce novel sequences at scale, and the field is accelerating. The AI protein design market is projected to reach $6.98 billion by 2033. But AI-designed proteins are not optimized for manufacturability the way traditionally engineered sequences are. As Twist’s press release put it, researchers need their “AI-generated sequences manufactured exactly, without adding a step to alter the sequences for manufacturability.” Those unoptimized sequences are more likely to contain the structural features, such as long repeats, inverted repeats and extreme GC content, that historically pushed orders into the “Not Accepted” bucket.

Complex sequences also come with a tradeoff in turnaround time. Twist’s Express service, which delivers standard clonal genes in four to seven business days, does not extend to complex orders. Those follow the company’s standard 15-business-day timeline. The specialist tailor, in other words, can take on the difficult job, but not on a rush schedule.

Twist is not alone in targeting the upper end of the market. Integrated DNA Technologies, one of Twist’s primary competitors, announced a partnership with Elegen in March 2025 offering clonal genes up to 15,000 base pairs with turnaround as fast as 10 business days. IDT also partnered with Ansa Biotechnologies for additional clonal DNA products.

In any case, the expanded service offerings reflect a widening gap between what computational protein design can propose and what commercial DNA synthesis can deliver. AI systems can now generate therapeutic-antibody candidates in minutes. But those designs still have to become physical molecules before they matter in the lab. For complex constructs, the DNA that encodes them can mean weeks of synthesis and QC, if a provider will accept the order at all. Even the most promising AI-generated candidate is inert data until someone can synthesize and test it. “It really truly isn’t useful unless you can put the molecule into a well and test it,” Finn said. Protein-design models score candidates on binding, stability and structure.  “The AI is optimizing for the biological performance of the feature rather than manufacturability.”

Certain sequence features have long posed challenges for commercial-scale synthesis. Elements like short tandem repeats, inverted repeats, long repeats and extreme GC content (the proportion of guanine and cytosine bases in a sequence, which at very high or very low levels creates secondary structures that stall synthesis) often appear in sequences tied to nucleic acid therapeutics, including cell and gene therapies and mRNA-based treatments.

Twist built complex-gene synthesis internally, extending its existing phosphoramidite chemistry platform with longer oligonucleotides (500-mers) and improved assembly workflows. “Being able to write longer oligonucleotides has been the molecular enabler,” Finn said. “You can use phosphoramidite chemistry to synthesize the complex part of the gene or construct de novo, and then stitch together the less complex parts to make the overall product.”

IDT chose a different path: reselling specialized partners’ capabilities through its own storefront. Both are pursuing the same one-stop-shop logic, but IDT’s partners are using fundamentally different chemistry. Ansa Biotechnologies uses an enzymatic synthesis process that achieves stepwise yields above 99.9%, compared to roughly 99.5% for phosphoramidite. That gap compounds at length. Ansa’s IDT partnership covers constructs up to 50,000 base pairs, more than seven times Twist’s new ceiling. For clonal DNA in the range comparable to Twist’s catalog (up to 7.5 kb), Ansa offers a nine-business-day turnaround with an on-time guarantee. Elegen, IDT’s other partner, reaches 15,000 base pairs in as fast as 10 business days.

Twist’s launch also coincides with its deepening role as wet-lab infrastructure for AI-driven drug discovery. In April, AWS announced Amazon Bio Discovery, an AI-powered platform that connects computational antibody design with lab synthesis and testing in a continuous loop. Twist, alongside Ginkgo Bioworks, is a core AWS lab partner. In the platform’s first publicized case, researchers at Memorial Sloan Kettering used it to design nearly 300,000 novel antibody molecules and sent the top 100,000 to Twist for synthesis and testing.