DT-7012 is a differentiated humanized anti-CCR8 IgG1 monoclonal antibody engineered for optimized binding and enhanced effector functions to deplete CCR8+ tumor-resident Tregs, while maintaining peripheral immune integrity. We will highlight the capacity of DT-7012 to preserve its binding to all forms of the CCR8 receptor, its antagonist and killing activities in the presence of competing CCL1, and the Fc-enhancing strategy of DT-7012. DT-7012 has entered a Phase 1 clinical evaluation, in participants with advanced solid tumors (DOMISOL; NCT06819735) and is entering a Phase 1 clinical evaluation, in patients with relapsed or refractory CTCL (CITY; NCT07213882).

G-protein-coupled receptors (GPCRs) transduce signals through the cell membrane by employing a catalytic reaction with intracellular GTP-binding proteins. We have demonstrated that this catalytic activity is reversible, and this reversal permits the GPCR to convey signals, across the cell membrane, in a complex manner. Using classical methods in pharmacology, we have investigated this mechanism at the mu opioid receptor and observe that agonists with GTP-release selectivity are able to prolong the antinociceptive effects of more conventional opioid agonists. Based on these findings, this form of agonist selectivity may provide a novel form of physiological signaling.

The angiotensin II type 1 receptor (AT1R) is a G-protein-coupled receptor (GPCR) targeted by frontline hypertension therapeutics. We and others have developed peptide and nanobody AT1R ligands with a range of biased signaling profiles. Here, we demonstrate that biased ligands can activate AT1R through multiple molecular mechanisms and thereby further diversify transducer activation. Leveraging these mechanisms could enable finer control of GPCR drug pharmacology than previously thought possible.

CXCR3 plays a central role in immune trafficking, inflammation, and cancer, but the mechanisms underlying differential activation by CXCL9, CXCL10, and CXCL11 remain incompletely understood. Using cryo-EM, pharmacological analyses, mutagenesis, chimeric chemokines, and molecular dynamics, we define distinct ligand recognition and signaling mechanisms. These results uncover key determinants of CXCR3-ligand multispecificity and signaling plasticity, providing a framework to better understand and modulate CXCR3 biology.

Proximity labeling has emerged as a powerful method for rapid and temporal mapping of transient and fragile protein-protein interactions, especially those in membranes. We exploit the photocatalytic activation of endogenous amino acids in proteins, enabling crosslinking with neighboring proteins for contact-dependent antibody mapping (contactMAP). This converts transient interprotein recognition events into covalent molecular records. We demonstrate the applicability of contactMAP for mapping high resolution snapshots of the Her2 and EGFR interactomes.

We recently announced JAM-2, our model for designing antibodies with drug-quality properties at high success rates. In this talk, we present results across challenging target classes, including GPCRs, pMHC complexes, and multipass membrane proteins, where JAM-2 frequently achieves zero-shot antibody design with single-to double-digit percent binding rates and strong affinities. We also discuss where the model falls short and what those failures reveal.

Discovering antibodies that functionally modulate GPCRs is a major challenge in therapeutic engineering. We combine high-throughput microfluidics-based functional screening with AI model training to generate and learn from large-scale functional datasets, enabling antibody design guided by true biological activity rather than binding alone. This integrated approach creates a scalable framework for engineering functional anti-GPCR antibodies with tailored modulatory properties.

GPCR antibody discovery is hindered by receptor instability, poor solubility, and limited access to native extracellular conformations. We use generative AI-based protein design to build soluble GPCR surrogates that display epitopes in stable, screening-ready formats. These engineered antigens reproduce selective binding with endogenous and pharmacological ligands. Using this platform, we enable single-experiment discovery of both species cross-reactive and receptor-subtype cross-reactive antibodies against challenging amylin receptor targets.

Membrane proteins remain difficult targets for antibody drug discovery and optimization. Adimab's yeast-based platform, paired with immune repertoires from llamas and humanized mice, enables the efficient generation and engineering of both HCAb and IgG antibodies. Using approaches such as nanodisc-based affinity selections from full-length yeast IgG and HCAb libraries, we demonstrate improved functional potency against membrane targets including the T cell receptor complex and GPCRs.

Recent clinical data show that antibodies against the P. falciparum circumsporozoite protein (CSP) can protect against malaria. This talk will discuss CSP as an example intrinsically disordered antibody target. CSP’s disordered central repeat region presents several related peptide motifs. Peptide cross-reactivity and the highly disordered nature of the central repeats present unique engineering challenges. Here we will explore a variety of approaches to discover functional antibodies against this important target.

Gut tissue damage mediated by inflammatory T cells circulating between the lymph nodes and intestinal tissues in Crohn’s disease, is currently treated by antibody or small molecule approaches targeting the integrin α4β7. We have developed an alternative, and potentially complementary approach that effectively inhibits T cell trafficking to the gut with a peripherally restricted & highly selective antagonist of the cannabinoid G protein-coupled receptor CB2R. Discovery, mechanism of action and preclinical efficacy data will be presented here.

For decades, attempts to identify TROP2 antibodies with native anti-tumor activity have been unsuccessful, and industry resorted to ADCs. Using AI-driven paratope mapping, KisoJi identified distinct antibody clusters not previously seen. From these clusters, KisoJi identified naked antibodies with potent anti-tumor activity. Further refinement using paratope barcoding showed >90% of antibodies sharing a barcode retained binding and >80% showed competitive overlap. KisoJi’s naked TROP2 antibody will enter clinical trials 2026.

No agent targeting a chemokine receptor has succeeded as an immunotherapeutic drug. One possible reason is the expression of a given receptor on leukocytes with opposing activities, so that receptor antagonists block the infiltration by both pro- and anti-inflammatory cells. I will describe a new method for cell-type selective inhibition of CCR2, a widely expressed receptor mediating migration of both pro-inflammatory CD4+ T (and other) cells and anti-inflammatory myeloid cells.

T cell dependent bispecific antibodies (TCBs) redirect cytotoxic T cells to tumor antigens, yet traditional assays lack kinetic insight and throughput. We developed an automated, high-throughput platform for multi-parametric assessment of TCB-mediated killing and cytokine production. Applied to hematological and solid tumors, the platform utilizes advanced data visualization to differentiate TCBs and characterize combination therapies. This novel approach enables robust functional profiling, enhancing the identification of optimal lead therapeutic candidates.

Due to their critical functions in cell interactions and signal processing, membrane proteins are often identified as important therapeutic targets. However, major limitations exist in developing antibodies that recognize complex, multipass transmembrane proteins. Here, we developed a comprehensive workflow, termed biofloating, that employs new suspension cell-based whole cell selection methodologies to screen for specific antibody binders against GPCRs utilizing naïve yeast-displayed antibody libraries. Binders were discovered against four disease-relevant GPCRs.

Agonist antibodies specifically targeting metabolic class B GPCRs could reduce the GI side effects that limit current obesity drugs. Abalone Bio’s FAST platform measures sequence-activity relationships of millions of antibodies to uniquely use ML for agonist discovery. With FAST, we have successfully identified and optimized agonist antibodies for a validated metabolic target and demonstrated efficacy with reduced side effects in obesity animal models.

This presentation will explore a nanodisc-based platform engineered to enable robust identification of antibodies targeting membrane proteins, with a particular focus on HLA complexes. We will also discuss how this platform enables the detection of donor-specific antibodies (DSAs) in serum from kidney transplant recipients, addressing a critical need in transplant immunology. Finally, the presentation will highlight how this technology can be extended beyond diagnostics to support therapeutic antibody discovery.

G protein–coupled receptors (GPCRs) and other multi-pass membrane proteins remain challenging targets for antibody discovery due to purification approaches that often disrupt native conformation and compromise stability, as well as limited extracellular epitope availability. Here, we present a virus-like particle (VLP)–based platform that enables robust presentation of membrane proteins in a native conformation. When combined with a synthetic VHH yeast display library and an optimized selection strategy, this approach enables efficient discovery of high-affinity binders. Collectively, this integrated platform expands the accessibility of difficult membrane protein targets for antibody discovery and therapeutic development.

A major challenge in AML immunotherapy is the lack of truly selective surface antigens. Using structural surfaceomics, our UCSF team identified the active conformation of integrin ß2 as an AML-specific target absent from normal hematopoietic progenitors. Building on this discovery, I am developing structure-guided, conformation-specific ADCs and cryo-EM-informed engineering strategies to improve selectivity, potency, and therapeutic index for AML treatment. This work focuses on optimizing linker chemistry, payload choice, and conjugation design, while also mapping antibody–antigen interfaces to guide affinity maturation and specificity. Together, these studies aim to advance a new class of precision immunotherapies for AML.

EGFR TKI resistance in lung adenocarcinoma (LUAD) involves extensive surfaceome remodeling. We developed a platform detecting surface protein conformation changes (SPCs), beyond expression alone. Patient-derived, EGFR-mutant LUAD models (TKI-sensitive and -resistant) were screened for SPCs; a deep learning framework ranked first-in-class targets by surface localization, altered protein-protein interactions, conformational change, and therapeutic relevance. Applied to Osimertinib resistance, this identified novel SPCs defining targets for next-generation ADC therapies in resistant NSCLC.

Biologics targeting GPCRs have historically been difficult to discover. Skape Bio builds GPCR modulators by combining AI protein design with high-throughput screening in human cells, and has discovered modulators for more than 10 GPCRs. We design agonists and antagonists with atomic precision and run rapid design-build-test cycles to deliver validated hits. This talk presents case studies across class A and B GPCRs and highlights scalable discovery of novel miniprotein therapeutics.

Advances in transferrin receptor 1 (TfR1)–mediated transcytosis are reshaping strategies for delivering large therapeutics across the blood–brain barrier. Systematic engineering of diverse TfR1-binding antibodies has revealed distinct design spaces that govern peak versus sustained CNS exposure and clarified how affinity, valency, and binding geometry influence trafficking, receptor turnover, and safety. These insights establish foundational principles for constructing optimized shuttles that enable next-generation biologics and genetic medicines targeting the central nervous system.

Brain delivery of diverse therapeutic cargos for treating neurodegenerative diseases requires a modular platform. Here, we engineered an antibody Fc-domain, called the TransportVehicle (TV), to bind highly expressed receptors for brain transport, avoiding non-native domains and maintaining peripheral half-life. TVs were engineered to bind either the transferrin receptor (TfR) or CD98hc. Both enhance brain delivery, but they exhibit differentiated kinetics. We also explore the combination of TfR and CD98hc binding in a Dual TVs approach to achieve higher brain concentrations than a single target alone. Modulation of TfR and CD98hc affinity independently can “tune” the molecule for the desired delivery properties.