For years, antibody discovery has largely revolved around one question: Can the molecule bind its target and produce the desired biological effect? Increasingly, researchers are asking another question much earlier in the process: Can that antibody actually become a drug?
Maria Gonzalez-Pajuelo, cofounder and Executive Vice President of Technology and Scientific Office at FairJourney Bio, told DDN that developability has become a much more important consideration during antibody discovery, rather than an issue left until later-stage development.
“It’s not only looking for a functional antibody but also asking whether it’s possible to make it,” she said. “It doesn’t help to have an antibody that binds and performs its function if, at the end of the day, you can’t put it into patients.”
For an antibody to progress from discovery to the clinic, researchers need to look at much more than just binding capability. They must also consider properties such as stability, solubility, expression, purity, and manufacturability. These characteristics can influence whether a candidate can be produced consistently, formulated for its intended route of administration, and developed on a practical timeline.
Historically, many of these questions were addressed much later during chemistry, manufacturing and controls (CMC) development. Moving them earlier in the discovery process could help researchers identify problematic candidates before substantial resources have been invested in their development.
Bringing developability upstream
Antibody discovery can generate hundreds or thousands of molecules that recognize a particular target. Researchers then progressively narrow this pool based on characteristics such as affinity, specificity, and biological function. But a molecule that performs strongly in a biological assay is not necessarily the easiest to develop.
Antibodies can have physicochemical properties that create problems later in development, including poor stability, low expression, aggregation, or difficulties with formulation. Identifying these issues after a candidate has already entered more advanced development can turn a promising lead into an engineering project, requiring researchers to go back and modify the molecule before it can progress.
The earlier these properties are assessed, the more opportunity there is to select candidates with favorable characteristics or address potential problems through engineering. “We need to make sure that what comes out of discovery is something that CMC can take forward easily,” Gonzalez-Pajuelo said. “If we don’t think about them at the discovery stage, they can become a bottleneck.”
That emphasis on the end product also affects decisions much earlier in the process, including how researchers choose to find their antibodies.
Matching discovery strategies to the drug
Antibody discovery has evolved into a broad toolbox of in vivo and in vitro approaches, from hybridoma technology and transgenic animals to phage, yeast, and mammalian display. Each method offers different advantages, so the most appropriate approach depends on the biology of the target and what the eventual therapy needs to do.
“Not one strategy fits all. Having access to all these technologies is wonderful, but we need to be mindful of what will be helpful in each situation,” Gonzalez-Pajuelo said. “It’s a combination of knowledge and being aware of the biology and the target product profile.”
The target product profile describes the characteristics a therapy will eventually need, including how it will work, how it will be administered, and the properties needed for its manufacture. Considering these requirements early can help researchers choose a discovery approach that is suited not only to finding a binder, but to producing a molecule that can ultimately become a drug.
It doesn’t help to have an antibody that binds and performs its function if, at the end of the day, you can’t put it into patients.
—Maria Gonzalez-Pajuelo, FairJourney Bio
That choice has become more nuanced as antibody therapeutics have expanded beyond conventional monoclonal antibodies to include bispecific and multispecific formats. The discovery strategy now needs to account for greater complexity, including how the antibody will function in its eventual molecular format and whether it has properties compatible with development.
During The Antibody Series 2026 conference, Gonzalez-Pajuelo advocated for the use of VHHs as building blocks for more complex antibody architectures. Their small size and modularity facilitate design flexibility, allowing researchers to tailor molecules precisely to the target product profile, enhancing potency, stability, and manufacturability.
“You don’t have to worry about pairing with another chain, and their smaller size is helpful when you’re building increasingly complex architectures. I think that helps a lot with bispecifics and multispecifics,” she said.
AI could accelerate optimization
The same emphasis on earlier optimization is also shaping how researchers use computational tools. Models can help identify antibody variants with improved affinity, specificity, or developability, which can then be tested experimentally through iterative design-build-test cycles. This could reduce the number of experimental rounds needed to arrive at a candidate with the properties required for development.
But AI's role in discovering entirely new antibodies is less settled. “I think everybody agrees that AI will speed up antibody engineering and optimization,” Gonzalez-Pajuelo said. “But when it comes to discovering novel antibodies, I think that’s still controversial. We probably need another five years to really understand what is possible.”
One limitation is the data available to train these systems. AI models can learn from large collections of antibody sequences and experimental results, but successful experiments are much more likely to be recorded than failures. Gonzalez-Pajuelo said the field needs more data, including negative data, to understand which approaches do not work.
For now, she sees AI as another tool that can complement established discovery and engineering approaches rather than replace them. “AI, automation, and all these tools will help us shorten the timelines,” she said.
As discovery technologies, antibody formats, and computational tools continue to evolve, the challenge will be turning a great antibody into a great drug — one that can be manufactured, formulated, engineered into the right format, and administered to patients.
