Multispecific antibodies (MsAbs) are engineered molecules that exhibit extensive structural diversity, a critical feature for biologic therapeutics. However, this complexity poses significant challenges in rational design and manufacturing. Here, we present an object-based encoding system that enables an agnostic and universal design framework for MsAbs, independent of predefined structural constraints. In addition, we are developing an integrated machine learning (ML) and molecular dynamics (MD) model to predict the developability profiles of full-length MsAbs, providing insights into stability and manufacturability. This toolbox, designed by protein engineers for protein engineers, represents a paradigm shift in MsAb design. By leveraging in silico approaches, scientists can rapidly generate, evaluate, and optimize thousands of MsAb formats before experimental validation, streamlining the discovery and development pipeline.

Brain delivery of therapeutics is highly challenging. TXP1, a single-domain anti-TfR1 antibody, enhances antibody brain delivery >40-fold in NHPs. Its brain selectivity stems from a unique binding epitope, while bispecific engineering ensures an unparalleled safety profile, even with full effector function antibodies. TXP1 marks a breakthrough in achieving high brain penetration, specificity and safety, offering hope for patients with CNS disorders.

T cell-engaging (TCE) multispecific antibodies demonstrate clinical efficacy but their use is partly limited by the small number of available anti-CD3 effector antibodies. This talk presents the discovery and engineering of novel anti-CD3 heavy chain-only antibodies (HCAbs), which exhibit T cell cytotoxicity similar to clinically validated TCEs and thereby provide a versatile new option for this potent class of biologics.

ImCheck developed bispecific antibodies with varied formats and valency to modulate anti-BTN3A agonist potency. These approaches enhance Vγ9Vδ2 T cell stimulation, block immune checkpoints, and explore cis/trans anchoring for stronger anti-tumor activity.

• Can molecular sampling identify surface patches that predict molecule developability?
• Can we achieve the speed and computational efficiency needed to process hundreds of full-length molecules using conformational sampling within a day or less?
• Can machine learning–based sampling methods replicate the conformational space of all-atom molecular dynamics (MD) for Building Blocks with at least 80% accuracy?
• What is the computational cost of running these models?​

We have developed novel immunocytokine format designed to prevent systemic immune activation and thus side effects. Leveraging unique properties of IL-15, we developed constructs where activity of the cytokine moiety is dependent on target binding by the antibody part. Based on this platform, we generated and characterised immunocytokines with favorable developability as well as high efficacy and safety for treatment of lymphoid malignancies and AML.

The concept and design of Proximity-Activated Cytokines (PACs) will be presented, using ANV700, a PD-1-targeted IL-21 PAC for the treatment of solid tumors, as an example. Through structure-guided design, an optimised anti-IL-21/IL-21 fusion protein was developed to enable selective signaling on PD-1+ T cells, thereby reinvigorating tumor-reactive T cells while minimising systemic cytokine exposure.

Cytokines emerged as promising molecules for therapeutic intervention to modulate the immune response. However, their often pleiotropic nature, combined with their high potency when administered systemically, restricts their therapeutic applicability. We have generated cytokine mimetics with tailor-made mode-of-actions based on multifunctional antibody derivatives. In this talk, cytokine mimetics for a rare disease indication will be presented as well as immunocytokine mimetics that specifically agonize TNFR1 for preferential cancer cell death induction.