When AI meets Phage Display: Accelerating the future for life-saving antibodies

Imagine trying to find one key that opens a single lock among billions of almost identical keys. That is remarkably similar to the challenge scientists face when searching for antibodies, the proteins our immune system uses to recognize viruses, bacteria and diseased cells. For more than 30 years, a technology called phage display has helped researchers solve this problem, leading to life-saving medicines for diseases ranging from cancer to autoimmune disorders. Now, a new partner has entered the picture: artificial intelligence (AI). Together, AI and phage display are transforming how antibodies are discovered, making the process faster, smarter and more efficient than ever before.

Phage display may sound complicated, but the idea is surprisingly elegant. Scientists use harmless viruses that only infect bacteria, known as bacteriophages or simply phage. Each phage carries the genetic instructions needed for its replication. When creating a phage display library, different antibody genes are inserted into the bacteriophage gene, resulting in the display of antibody fragments on the surface of the phage, typically fused to a gene III coat protein.  

By creating billions of unique phage, each displaying different antibody fragment, researchers create enormous libraries. These libraries can be screened to identify antibodies that bind to disease-related targets, with potential applications in treating cancers, autoimmune diseases, inflammatory and viral infections.

This approach has revolutionized drug discovery. Several approved antibody drugs, including treatments for inflammatory diseases and cancer, originated from phage display technology. One notable example is Humira, which was developed using phage display technology and became the world’s best-selling medicines for rheumatoid arthritis and other inflammatory conditions. Yet the process remains demanding. Although researchers begin with billions of antibody candidates, only a small number possess the combination of strong binding, stability and safety needed to become a medicine. Identifying these rare candidates often requires extensive laboratory screening, contributing significantly to the time and cost of drug development.

This is where artificial intelligence is beginning to make a real difference. Much like AI systems that learn to recognize faces, translate languages or generate text, modern AI can also identify complex patterns in biological data. By analysing millions of antibody sequences collected from humans and other species, AI can recognize the characteristics that make some antibodies more effective than others. Rather than searching blindly through a vast number of candidates, scientists can now use AI to predict which antibody designs are most likely to succeed before they are ever built and tested in the laboratory.

One of the biggest opportunities lies in designing better antibody libraries. Traditionally, researchers generated diversity by introducing random changes into antibody genes or by incorporating rational sequences based on approved antibody biobank, hoping that some combinations would produce useful molecules. AI can be used in two parallel campaigns. First, it can design a pool of antibodies that are more likely to recognise specific targets while maintaining antibody structural integrity. Second, AI can assist after the laboratory experiments begin. Modern phage display experiments generate enormous amounts of DNA sequencing data, revealing how millions of antibody candidates perform during the selection process. AI can rapidly analyse these datasets, identify the most promising antibody families and even distinguish genuinely strong binders from those that appear successful for technical reasons. This can be either used to refine future library designs or help researchers focus on the most promising candidates much earlier in the discovery process, saving both time and resources.

AI is also changing what happens after an antibody has been discovered. Not every antibody that binds well in the laboratory becomes a successful medicine. Some are difficult to manufacture, while others may be unstable or trigger unwanted immune responses. AI models can now predict many of these properties before expensive testing begins, helping researchers prioritize the antibodies most likely to become safe, effective treatments and commercially viable treatments.

Perhaps the most exciting aspect of this partnership is that AI is not replacing laboratory science—it is strengthening it. Phage display provides the experimental proof that an antibody works, while AI helps scientists make better decisions at every step of the discovery process. Each experiment generates new data that improve AI models, and better AI models help design more effective experiments. Together, they create a powerful cycle of learning and discovery.

The potential impact extends far beyond today's medicines. Researchers hope these technologies will speed the development of antibodies against emerging viruses, difficult cancers and autoimmune diseases, while eventually enabling more personalised treatments tailored to individual patients. Challenges remain, including ensuring the accuracy and interpretability of AI predictions and avoiding excessive reliance on computational models. Nevertheless, the one thing that is becoming clear: the future of antibody discovery is likely to be driven not by artificial intelligence alone, or by laboratory experiments alone, but by the combination of both.

As AI continues to reshape many areas of science, its partnership with phage display offers an excellent example of technology augmenting rather than replacing human expertise. The next generation of life-saving antibody medicines may emerge not only from the ingenuity of scientists at the laboratory bench, but also from intelligent algorithms working alongside them.