Single-domain antibodies (sdAbs) are compact, stable, single-chain biologics increasingly used in diagnostics, radio-immunoconjugates, and multispecifics. As a newer therapeutic format, the developability rules for sdAb containing Biologics are still emerging. We share lessons from the comprehensive biophysical profiling of >30 clinical-stage sdAbs in the VHH-Fc and VHH formats, spanning developability and manufacturability metrics, alongside preliminary in vivo PK results.

Enteric diseases are common human ailments. However, conventional biologics are not suited for enteric applications due to their susceptibility to degradation in the protease-rich hostile environment of the GI tract. Using click display, which directly links the protein library to the coding cDNA, and a language model to access "naturalness," we report the engineering of protein-stable DARPins for potent neutralization of C. difficile toxin TcdB as oral therapeutic candidates.

The modest ability of antibodies to penetrate the blood-brain barrier severely limits their use in therapeutic applications. We are developing antibody shuttles that target CNS proteins to mediate enhanced and selective CNS targeting and, in some cases, long-lived CNS retention. Here we will discuss our recent progress in engineering next-generation transferrin receptor and CD98hc-targeted CNS shuttles, as well as their application for delivering proteins and nucleic acids for therapeutic applications.

The blood-brain barrier presents a major obstacle to brain drug delivery. We have developed an enabling platform for the identification of blood-brain barrier targeting antibody-like molecules known as Variable Lymphocyte Receptors (VLRs). These VLRs could ultimately be used to ferry drug cargo into the brain. Here we will describe our recent efforts to identify and validate such blood-brain barrier targeting VLRs.

Targeting blood–brain barrier receptors enables brain delivery of biologics. We engineered an Fc-based dual transport vehicle (TV) that targets transferrin receptor (TfR) and CD98hc, combining rapid TfR-driven uptake with CD98hc-mediated retention. Dual TVs achieve higher brain concentrations than single-receptor formats. Adjusting TfR/CD98hc affinities tunes exposure kinetics and biodistribution. A mechanistic model links architectural designs to brain PK, guiding optimization of brain-penetrant biologics.

Accurately predicting developability characteristics is a pivotal yet challenging task in antibody therapeutic development. To overcome the limitations of small training datasets, we designed and implemented an integrative antibody feature engineering and machine learning framework called PROPERMAB. Using this framework, we developed predictive models for antibody hydrophobic interaction chromatography retention time and high-concentration viscosity. We also demonstrate the potential to scale our approach to repertoire-scale sequence datasets.

Experimental binding affinity data offer an untapped opportunity to train AI models for improved antibody design. We present an AI model that combines structural representations with experimentally measured affinities to guide sequence generation. Compared to the general-purpose protein language model, our model demonstrates enhanced ability to generate antibody variants with higher predicted binding affinity in silico. Early experimental evaluations further highlight the complex relationship between affinity and developability, underscoring the importance of integrating multi-objective optimization strategies to advance AI-guided antibody engineering.

Cytokines are vital for anti-tumor immunity but face several limitations as therapeutic molecules, including pleiotropic activity, poor stability and half-life, and dose-limiting toxicity. Different engineering approaches to modify the cytokine or turning to alternative formats such as antibody mimetics exist to address these limitations. In this talk, we describe engineering approaches undertaken to develop cytokine Fc fusion and mimetics for IL-27 agonism.

Antibody developability is difficult to ensure using traditional phage or yeast display platforms, which often yield binders with poor biophysical properties compared with those generated by animal immunization. To overcome this, we developed a mammalian selection platform based on pseudotyped lentiviral display that enables direct on-cell selection of antibodies in their native folding and secretion environment. This approach generates antibodies with markedly improved developability and accelerates discovery against challenging targets.

Voltage-gated sodium channels initiate action potentials in neurons and muscle. Dysfunction of these channels leads to sustained sodium influx that underlies various human diseases including cardiac arrhythmias and epilepsy. We used de novo protein design to engineer a peptide modulator that restores the proper function of these ion channels. These studies demonstrate the therapeutic potential of rationally designed biologic agents for correcting ion-channel dysfunction across cardiac and neurological disease settings.