Main Conference Day 3Wednesday, 17 December 2025 - PT (Pacific Time, GMT-08:00)

My group engineers genetic systems that dramatically accelerate the speed of mutation and gene evolution in vivo so that we can drive the rapid evolution of new biomolecular functions and prospectively watch (and systematically manipulate) the course of long gene evolutionary processes on laboratory timescales. I will share recent developments in the use of our continuous evolution system, orthogonal DNA replication system (OrthoRep), to evolve antibodies. I will discuss our efforts to affinity mature antibodies at scale along with the intersection of computational antibody design and evolution, including work focusing on prioritizing sequence space exploration to generate data for training computational models.

We developed a novel functional screening method using hyperphage display platform that allows rapid discovery of potent antiviral single domain antibodies. We benchmarked RASP against established phage ELISA and deep sequencing methods. RASP can be used either as a standalone platform or seamlessly integrated with conventional screening methods to accelerate the discovery of antiviral VHHs.

We present JAM-2, a fully computational antibody design platform that produces VHH-Fc and full-length mAbs with drug-like affinity and developability. Across 16 unseen targets and in under a month, JAM-2 achieved 100% target coverage with double-digit percent bind rates and frequent picomolar to single-digit nanomolar binders from just tens of designs tested. JAM-2 also directly generated GPCR-targeting antibodies in native cellular contexts, delivering lead-quality molecules without post-hoc optimization.

Therapeutic protein engineering has been transformed by the incorporation of big data and AI/ML techniques. An emerging challenge for this field is how to efficiently leverage the right data and the best models to drive meaningful results and resolve long-standing bottlenecks. Amgen has incorporated a generative biology approach to tackle complex engineering problems, aiming to deliver better, more effective molecules across every therapeutic program.

This talk will share updates from the AIntibody competition, a benchmarking initiative engaging the biotech, pharma, academia, and AI communities to use AI and other informatic methods to design or identify developable antibodies with high affinities, from curated NGS datasets. Results will compare the properties of these antibodies with those derived using experimental methods, providing insights into the value of AI in antibody discovery. AIntibody announcement manuscript: Erasmus, M. F. et al. Nat Biotechnol 42, 1637-1642 (2024).