What we learned at PEGS 2025
Highlights from the Frontlines of Antibody Discovery and Engineering
Read Time: 5 minutes
PEGS 2025 brought together platform innovators, AI disruptors, and biologics leaders shaping the next generation of antibody therapeutics. From smarter ADCs to AI-designed antibodies and single-cell screening breakthroughs, the momentum is real and accelerating.
ADCs are Having a Renaissance
One of the major themes of the conference was the accelerating progress in antibody-drug conjugates (ADCs). AstraZeneca’s Puja Sapra presented compelling data in the plenary keynote showing how small changes to linker chemistry or payload choice can significantly improve the safety and efficacy of HER2-targeted ADCs. Trastuzumab deruxtecan (T-DXd), which uses a cleavable linker, outperformed Trastuzumab emtansine (T-DM1) in both HER2-high and HER2-low tumors, as shown in this Nature study, highlighting how nuanced chemical engineering can redefine clinical efficacy.
This level of sophistication is now extending into multispecific ADCs and site-specific conjugation using non-natural amino acids. AstraZeneca’s DuetMab platform exemplifies this trend, enabling tunable affinity in bispecific formats that can reduce toxicity while improving target engagement.
These innovations underline the need for functional screening and rapid profiling across large panels of antibodies, something we help facilitate through our rapid discovery workflows and flexible assay development capabilities.
AI in Antibody Design: Hype, Hope, and Hard Data
Artificial intelligence remains a hot topic, but the focus is shifting from potential to performance. Several groups shared real-world applications where AI models are not just assisting but actively driving antibody discovery and development.
MOLCURE demonstrated how AI-selected antibodies from a VHH phage library reached picomolar affinities, and how LLM-generated antibody sequences found high-affinity binders outside the original diversity set. GV20 Therapeutics, meanwhile, leveraged immune repertoires derived from patient tumor data to identify rare, potent antibodies against previously unvalidated targets like IGSF8, a promising new immuno-oncology candidate.
Ginkgo Bioworks took a more foundational approach, showcasing their PROPHET Ab platform. It automates the generation of developability data — such as hydrophobicity (HIC), self-association (AC-SINS), and polyreactivity — at high throughput, producing AI-ready datasets with raw data access and metadata-rich outputs. This kind of data infrastructure is critical for training better models, and we’re excited to see how tools like this can complement empirical discovery efforts like our own, especially when screening large libraries or optimizing downselected leads.
Specifica shared the results from the first ever AIntibody Challenge, a blind head-to-head competition in AI-based antibody engineering. Participants (all blinded) received the same NGS datasets and were tasked with selecting, optimizing, and designing antibodies in three separate challenges. Interestingly, smaller academic and biotech groups outperformed big pharma and dedicated AI firms across most of the criteria. So far, the AI-based affinity maturation showed promising results, shaving 2–3 weeks off the traditional workflow. AI still struggles with selecting high-affinity antibodies from clonotype groups, and de novo design showed mixed success.
AI is a powerful tool, but still benefits from validation, diverse datasets, and iterative design.
High-Throughput Meets Single Cell
We saw several like-minded groups showcasing single B cell discovery workflows, an area where our AbTheneum™ platform has long been a pioneer. Sphere Bio's Cyto-Mine platform uses picodroplets to screen against antibodies in different multiplexed assays. OmniAb and Antibody Solutions presented workflows using xPloration in antibody discovery. We were interested to see the sequence diversity breakdown between plasma, memory, and activated B cells. Bruker Cellular Analysis showcased the Beacon Discovery applied to a viral target, implemented by La Jolla Institute for Immunology.
Having worked with antibody developers across pharma and biotech for over a decade, we’ve seen firsthand how powerful single-cell approaches can be and how often the right sequencing strategy makes all the difference. We have advanced methods for isolating more hits, multiplexing assays tailored to the target biology, and adapting to new host systems.
Most platforms follow a two-step process: screening, then selecting cells to sequence their antibodies. AbTheneum, by contrast, screens and captures all IgG sequences in parallel without selecting/picking, combining screening data with sequence data for more informed hit selection.
Case Study Cinemas: Our New Approach to Sharing Real-World Successes
We’re excited to announce that, at PEGS, we debuted a new way of sharing case studies: Case Study Cinemas. This innovative approach presents real-world success stories framed in the context of movies. Our goal? To make scientific storytelling more engaging and relatable while still providing actionable insights.
Conclusions
Whether it’s optimizing linkers for better ADCs, building antibody panels from patient repertoires, or refining AI pipelines with robust data, the message from this year’s conference was clear: the tools are getting better, and the timelines are getting shorter.
At Single Cell, we’re excited to contribute to this momentum — whether by enabling functional screening of full repertoires, tailoring discovery workflows for complex targets, or helping teams turn hits into leads faster. If you’re navigating your next antibody discovery campaign, whether it's a tough target or a tight timeline, we’d love to explore how we can help.