How AI and High-Throughput Library Generation Are Accelerating Antibody Discovery
Submitted by Emily Leproust | Twist Bioscience.
Originally published in Life Sciences Insights Magazine, August 2025
Therapeutic antibodies have a powerful impact on human health. These medicines are revolutionizing how we treat cancer, autoimmune diseases and many other conditions. Because antibodies can have such high affinity for their therapeutic targets, they offer tremendous promise to improve care.
Still, antibody drug development is a hard road. There are a massive number of potential molecules to choose from. The human body can create as many as 1020 distinct antibodies. To put that in context, there are roughly 1023 stars in the universe.
This impressive diversity underscores why antibodies play such critical roles in human immunity, but it also makes identifying suitable molecules incredibly challenging. Finding a needle in a haystack would be far easier.
Fortunately, AI is evolving into a powerful tool to accelerate and de-risk antibody drug development. Biopharma companies are actively developing AI tools that can help them narrow their choices, making it easier to identify potential therapeutics.
This is where Twist comes in: to translate digital to physical. Turning a sequence design into validated data at scale enables the ability to cull this massive universe of molecules to find the handful that are readily developable and possess therapeutic value. Conducting these studies in silico, rather than in the lab, dramatically accelerates our ability, as a global industry, to create valuable therapies for patients in need.
Identifying Therapeutic Antibodies
At Twist, we provide the enabling materials, including DNA, proteins, antibodies and data that help therapeutic development companies take potential treatments to the clinic. Now, increasingly critical in this evolving ecosystem, is our ability to generate high quality characterization data in very high-throughput.
Going back to the 1020 problem, one of the most powerful things we can do is help companies narrow that funnel. Because Twist can write DNA at scale, we can build incredibly large libraries that routinely exceed one trillion distinct sequences. As our customers use AI to generate antibody sequences, we can build libraries with the specific sequences that they need and generate both positive and negative training data for refinement of their AI tools. Furthermore, Twist can produce and generate data for thousands of distinct sequences arising from de novo designs, library screening, or any other output, including but not limited to in-depth binding and developability (Tm, Tagg, polyreactivity, etc.) characterization. Screening and drug selection at this scale was previously only possible for small molecules, but Twist enables this for large molecules (proteins and antibodies).
AI is not a magic wand that can solve all of biopharma’s antibody challenges. And cannot predict whether a specific antibody will become a successful drug. It also cannot provide straight yes or no answers about a molecule’s developability, but rather, can provide probabilities. Human judgment will always play a crucial role in these decisions. However, AI increases the number of high-quality shots on goal and becomes increasingly important to de-risk the decision making around whether an antibody has the necessary qualities to warrant further development.
Streamlining this workflow is having a profound impact on cost and throughput of the early stages of antibody discovery. We believe this process will ultimately shorten the time and increase the probability of success for taking a program from a concept to an approved medicine.
Providing the Data That Teaches Biopharma Models
To be clear, we do not currently build the computational models biotech and pharma companies use to delineate the best possible antibodies — we provide the data that informs those models.
We bring value in combining high-throughput protein and antibody expression, purification and analytics under a seamless process and business model. Because we can create and characterize tens of thousands of antibodies and have plans to expand this, we are poised with the ability to give the industry the tools to build their high-quality models.
In other words, we create the data that creates the model that creates the drug that addresses the disease that positively affects our goal of expanded global health.
Get the latest thought leadership from California’s life sciences sector in the quarterly Life Sciences Insights magazine. Share your own news and insights with CLS Member Voice.
FAQ: AI and High-Throughput Antibody Discovery
The search space is astronomically large. The human body can create as many as 1020 distinct antibodies — for context, there are roughly 1023 stars in the universe. That diversity is what makes antibodies so powerful in immunity, but it also makes identifying the handful of developable, therapeutically valuable molecules extremely challenging.
AI helps narrow the enormous funnel of possible antibodies, allowing companies to focus on the most promising candidates and de-risk development. Conducting much of this work in silico, rather than in the lab, dramatically speeds up the industry’s ability to create valuable therapies for patients in need.
Because Twist can write DNA at scale, it can build libraries that routinely exceed one trillion distinct sequences and produce both positive and negative training data to refine customers’ AI tools. This brings screening and selection at a scale once possible only for small molecules to large molecules like proteins and antibodies.
No. AI is not a magic wand — it cannot guarantee that a specific antibody will become a successful drug, nor give a definitive yes or no on developability. Instead, it provides probabilities and increases the number of high-quality “shots on goal,” while human judgment remains crucial to the final decisions.
Twist does not build the computational models companies use to select antibodies; it provides the data that informs those models. By combining high-throughput protein and antibody expression, purification, and analytics, Twist can create and characterize tens of thousands of antibodies, giving the industry the tools to build high-quality models.