TAVO412 was a trispecific EGFR x cMET x VEGF antibody with F243L/R292P/Y300L/V305I/P396L and knob-in-hole mutations. The molecular construct was designed with dual EGFR binding domains on the N-terminal and anti-VEGFA ScFv on the C-terminal of one heavy chain, and an anti-cMET Fab arm on the other chain. TAVO412 was expressed from a single stably-transfected CHO cell line, purified using Protein A and ion exchange chromatography, and characterized by SEC-HPLC and SDS-PAGE. We outline the selection process to obtain a stable molecule for clinical trials.

Intramolecularly-assembled immunocytokines (ICs) are fusion proteins that link an anti-cytokine antibody to a cytokine, circumventing shortcomings of recombinant cytokine therapy such as short half-life and off-target effects. To expand our IC platform, we designed dual-agonist immunocytokines to deliver two discrete signals to target cells. IC expression and production can be difficult due to assembly challenges and oligomerization; this talk highlights protein engineering strategies to enhance IC production for therapeutic development.

We have used codon deoptimization to improve bispecific protein expression and reduce production of mispaired species. While codon optimization avoids rare codon usage to accelerate protein translation, we found that deliberately incorporating non-optimal codons into certain heavy and light chain coding sequences can increase final product purity after a single-step Protein A purification. Our work suggests that rational codon usage may be a simple strategy for balancing chain expression.

Transient expression of complex bi- and multispecific antibodies can limit material supply for preclinical discovery and developability studies. We evaluated Leap-In® transposase technology to generate stable CHO pools expressing 1+1 and 2+1 common light chain bispecific antibodies. Vector engineering improved chain balance, heterodimer assembly, and productivity. The best pool reached 2.33 g/L, with 93.8% heterodimer and high purified antibody quality and target binding.

Using deep learning, we de novo designed recombinant proteins targeting short- and long-chain α-neurotoxins and cytotoxins while prioritizing properties that support efficient expression and production. The resulting proteins showed high thermal stability, strong binding affinity, and close agreement with computational models. They neutralized all three toxin subfamilies in vitro and protected mice from lethal neurotoxin challenge, demonstrating how computational design can enable stable, readily expressible next-generation antivenom proteins.

Recombinant AAV production is often limited by inefficient genome packaging and high levels of empty or partially filled capsids. We engineered AAV Rep proteins through directed evolution of a library derived from serotypes 1–13 in a producer cell line. Selected hybrid Rep variants showed substantial differences in packaging performance, with top variants improving AAV2 genome packaging efficiency by up to 2.5-fold, demonstrating a strategy to enhance rAAV vector production.

Seeking to accelerate the study of oxygen-sensitive proteins, we are working to develop an automated anaerobic cell-free protein synthesis (CFPS) platform. Through miniaturization and acoustic liquid handling, we can achieve much higher throughput at lower volumes compared to traditional expression methods. In this talk, I will describe our optimization efforts and present on our successful anaerobic CFPS of various proteins, including an enzyme containing an iron-sulfur cluster.

This presentation will highlight Immunethep’s strategy to advance a peptide-based vaccine concept targeting bacterial GAPDH (bGAPDH) to help prevent E. coli infection. We will review the target rationale and antigen design, including a tandem peptide construct built from selected bGAPDH-derived regions to target a conserved virulence factor while avoiding cross-reactivity with human GAPDH. We will summarize the development path, early learnings, and next steps toward a scalable, vaccine-ready profile.

Enzymatic macrocyclization offers a route to structurally complex peptides that are difficult to access through conventional methods. This presentation will describe recombinant production strategies that pair peptide expression with PapB-mediated thioether formation to generate conformationally constrained macrocycles and polymacrocycles. Key considerations include construct design, enzyme compatibility, reaction efficiency, product recovery, and analytical validation, with emphasis on enabling diverse peptide architectures for discovery and therapeutic development.

Machine learning and generative AI are transforming antibody discovery by enabling prediction of vast antibody sequence spaces, increasing demand for scalable experimental validation. AstraZeneca has reengineered its automated infrastructure to support high-throughput generation of transfection-ready constructs. This presentation will describe the strategic development and deployment of AstraZeneca’s end-to-end automation platforms to reduce cycle times, expand library generation, and enhance production and delivery capacity for simple to complex antibody formats, enabling validation of data-driven predictions at discovery scale.

Producing diverse therapeutic proteins efficiently requires more than a standardized workflow. This presentation will explore how reusable approaches across construct design, expression, purification, and characterization can be combined with molecule-specific decision-making. Drawing on experience leading protein science teams in therapeutic development, I will discuss how to recognize when platform methods are sufficient, when adaptation is needed, and how knowledge sharing across programs reduces repeated work. Attendees will gain practical perspectives on balancing standardization with scientific judgment to accelerate problem-solving and improve development efficiency.

Production of protein complexes and/or their constituents can be challenging given that the solubility and stability of the complex components may be dependent on their physical association with each other. Cellular localization (e.g. extracellular, membrane-anchored) and unique structural features (e.g. disulfide bonds, post-translationalmodifications) are additional complicating factors. AlphaFold-based structure prediction was used to guide construct design and prioritize candidate domains and putative subassemblies of a large extracellular microbial complex for experimental study. Conditions for large-scale expression and purification were determined by a 96-well expression/solubility screen.

Protein production workflows should be designed around the downstream question the reagent must answer. This presentation will explore how expression and purification choices influence protein quality, functionality, and suitability for structural, biochemical, and biophysical studies. Emphasis will be placed on defining fit-for-purpose requirements early and using characterization to confirm that produced proteins are suitable for their intended discovery applications.