AstraZeneca’s InSiDe (in silico Developability) platform provides cross-pipeline insights into antibody developability risk via machine learning models for non-specific binding, self-association, chemical liabilities, etc. As interest in multi-specific therapeutics grows, so does the need for in silico developability predictions for these complex formats, posing additional challenges relative to conventional antibodies. Here, we identify gaps where computational approaches used for conventional antibodies are insufficient and discuss approaches to overcome these hurdles.

We developed a robust peptide therapeutics discovery platform based on ~500 venom peptide scaffolds, integrating phage and yeast surface display technologies. Libraries were designed using machine-learning (ML) models that predict key residues for peptide foldability. An ML-enabled, rapid, and cost-effective affinity maturation workflow accelerates lead optimization, enabling the efficient identification of potent and stable peptide candidates against diverse therapeutic targets.

Yeast display platforms are foundational technologies that have long enabled critical advances in protein engineering and characterization. We will present recent innovations that expand their utility to new applications. We demonstrate these platforms can be powerfully leveraged to directly identify probes reactive to tumor metastasis, and to generate sensors targeting distinct post-translationally modified protein states. These advances open new horizons for precise therapeutic targeting of cancer and the immune system.

Peptide-HLA (pHLA)-directed T cell engagers can target otherwise inaccessible intracellular viral and tumor-associated antigens, but require both high affinity and high specificity for safe and effective clinical use. We present an integrated high-throughput yeast-based platform to rapidly discover, engineer, and de-risk fully human soluble T cell receptors (TCRs) and TCR-mimetic (TCRm) antibodies with high affinity and high selectivity, enabling the development of potent and specific pHLA-targeting therapeutics.

Antibody-small molecule hybrids integrate the molecular recognition capabilities of antibodies with chemical features (e.g., pharmacophores or covalent moieties). Screens with chemically expanded antibody libraries demonstrate that (1) a diverse set of hybrid architectures lead to inhibitors of carbonic anhydrases; and (2) pharmacophore properties can strongly influence screening outcomes. These findings highlight the importance of using an “experiment-first” approach to drive hybrid discovery toward clinically relevant applications. 

• The "discovery of the unexpected" remains valuable—balance with computational/ML approaches?
• What are best-use cases for adding chemistry to antibodies (or other polypeptides)?
• Biomanufacturing with noncanonical amino acids?

Bispecific proteins empower immune engagement, logical targeting, and pharmacokinetic modulation. Yet multi-domain antibody bispecifics require challenging engineering and limit delivery. We address this limitation by engineering a β-sheet in single-domain VHH antibodies as a second binding site. Dual-binding VHHs were engineered to multiple targets within multiple parental antibodies. Modularity is evidenced by grafting engineered β-sheets into other VHH, scFv, or IgG and incorporating a dual-binding VHH into a CAR.

Mammalian display libraries can be interrogated consecutively for developability and specificity. Similarly, libraries secreting antibodies are a perfect match for microfluidics-assisted high-throughput, function-first screens. The versatility and fruitful options arising from combining these emerging technologies will be discussed.

Native heavy- and light-chain pairing is increasingly recognized as important for therapeutic antibody function and developability. This talk describes a droplet-microfluidics platform that combines high-throughput single-cell encapsulation, molecular biology, and bioinformatics to recover cognate antibody genes and generate natively paired antibody libraries from antigen-specific B cells. Preserving native chain pairing yielded therapeutic antibodies with enhanced target specificity, potency, and developability compared with conventional combinatorial libraries, providing a scalable strategy for immune repertoire mining and the discovery of lead binders for antibody-based therapeutics.

Peripherally administered enzyme replacement therapies (ERTs) to address lysosomal storage disorders have been limited in their ability to target the CNS. Such ERTs are also limited by their instability in serum. To generate enzymes with sustained activity in serum, we developed a novel enzyme engineering strategy utilizing both yeast surface display and secretion technologies. We show durable activity with a POC enzyme in serum for many days. By fusing engineered enzymes to a TfR transport vehicle (TV) for delivery across the BBB, we also show improved substrate reduction in the brains of a relevant mouse disease model.