EpiTACs are bispecific antibodies in which one arm binds a pathogenic target, and the other arm leverages tissue-enriched degrading receptors to selectively degrade a wide range of extracellular targets including membrane, soluble, and multi-span proteins. Our modular and industrial process for creating EpiTACs allows us to optimize antibody properties to maximize degradation. EpiTACs to multiple oncology and autoimmune targets demonstrate that target degradation drives robust in vivo activity.

Antibodies targeting the conserved prehairpin intermediate (PHI) of class I viral fusion proteins are typically weakly neutralizing and not considered viable therapeutics. We enhanced the potency of an antibody targeting the transiently exposed gp41 N-heptad repeat (NHR) in HIV-1 PHI by engineering bispecific antibodies (bsAbs) that preposition it to the HIV-1 receptor or coreceptor. These bsAbs showed exceptionally broad neutralization. Despite sharing the same neutralizing arm, the bsAbs exhibited distinct resistance profiles. These findings validate the NHR as a therapeutic target and support the development of a new class of broadly neutralizing HIV-1 antibodies.

The inability of diverse biomolecules to readily penetrate the blood-brain barrier is a key limitation to their use in research, diagnostic, and therapeutic applications. We are developing bispecific antibodies that engage either CD98hc or transferrin receptor, and efficiently transport biomolecules into the CNS. We will discuss the unique advantages of each shuttling pathway, our progress in developing next-generation shuttles, and their drug-delivery applications.

BioCopy’s ValidaTe platform is a pHLA-centric workflow for developing safer and more effective T-cell receptor mimic (TCRm) antibodies. It combines proprietary ultra-high-throughput pHLA kinetic screening with AI-based predictions, structural modeling, and deep in vitro validation to comprehensively assess target specificity and efficacy. The platform’s strength is demonstrated by the discovery of novel, superior TCRms against the cancer-testis antigen MAGE-A4, showing an unprecedentedly minimized off-target profile.

Antibodies have become mainstream therapeutics due to their high precision, potency, and limited adverse effects. For neurological disorders and cancer, especially with poor prognosis, they represent highly attractive therapeutic agents, yet organ penetration remains challenging. We demonstrated that endovascular increasing osmotic pressure beyond current clinical standards followed by intra-arterial antibody infusion dramatically improves antibody extravasation to the target organs, including brain, in a safe manner, offering a promising therapeutic strategy.

This presentation explores a novel approach to enhance drug delivery by engineering bispecific antibodies that target the low-density lipoprotein receptor (LDLR). By leveraging LDLR's natural ability to rapidly internalize and recycle, we can improve the targeted delivery and cellular uptake of therapeutic payloads. This strategy provides a blueprint for developing next-generation bispecifics with enhanced efficacy in cancer treatment and other diseases requiring precise targeting.

T cell engagers (TCEs) are a promising approach for solid tumors, with the first TCE for a major solid tumor approved in 2024. However, progress is limited by a lack of tumor-selective targets that maximize efficacy while minimizing on-target, off-tumor toxicity. We developed the ATLAS platform, a curated single-cell RNA sequencing resource that leverages both healthy and tumor-derived datasets to identify highly precise TCE targets. Using ATLAS, we discovered LY6G6D as a tumor-specific colorectal-cancer (CRC) target, guiding the engineering of a best-in-class LY6G6D TCE for CRC.

Deep learning–based approaches have revolutionized protein structure prediction, yet the structural modeling and design of antibodies and novel protein interactions remain challenging. We present a physics-aware deep learning framework that addresses key limitations by incorporating biophysical principles into the learning and inference process. We demonstrate applications of this approach in combination with antibody engineering to design agonistic bispecific antibodies, as well as its integration with multi-omics data to uncover mechanisms of disease.

LabGenius Therapeutics’ EVA platform leverages avidity-driven selectivity to overcome T cell engager (TCE) challenges, including on-target, off-tumor toxicity in solid tumors. In this talk, we describe how the closed-loop integration of high-throughput experimentation with machine learning has facilitated the discovery and optimization of multispecifics for function and developability. Specifically, we showcase how we have developed a pipeline of TCEs that exhibit ‘on/off killing selectivity’ for targets with minimal expression differences.

The design of multispecific antibodies presents unique computational challenges due to the need to evaluate and optimize multiple binding interfaces simultaneously. In this talk, I will present a methodological framework that integrates structural models (e.g., AlphaFold), protein language models, and affinity prediction data across diverse platforms and targets. By combining these representations with graph-based machine learning, our models capture key determinants of binding. Our multi-layered approach supports mutation-effect prediction and affinity optimization, even with limited or noisy data. I will highlight how it improves generalization across antibody–antigen pairs and enables more rational antibody design.