Delivery of therapeutic payloads in a cell and tissue specific manner has given rise to several classes of novel medicines – from ADCs to oligonucleotide conjugates to targeted radiopharmaceuticals. The presentation will provide an overview of concepts and methods for selection and characterization of ligands enabling these delivery approaches.

Zelenirstat is the first N-myristoyltransferase (NMT) inhibitor to be tested in patients. NMT inhibitors work by multiple mechanisms to kill cancer cells. Phase 1 results with oral zelenirstat showed good safety and extended survival in solid tumor patients. As an ADC payload, zelenirstat was significantly more potent against solid tumor cancer cells using several different targeting antibodies and linkers. Dozens of related NMTis may also serve as ADC payloads.

Antibody–oligonucleotide conjugates (AOCs) and peptide–drug conjugates (PDCs) are novel therapies that combine targeted delivery with strong intracellular effects but introduce unique safety risks. AOCs may cause immunogenicity from new conjugation sites, activate innate immunity via toll-like receptors, or induce unintended gene modulation through off-target hybridization.  For PDCs, risks include proteolytic breakdown, fast renal clearance, off-target binding, and payload leakage leading to toxicity. Both face aggregation, heterogeneity, and narrow therapeutic windows.

I will present novel protein conjugates that can be activated by light. One example is photoimmunotherapy (PIT), in which the photosensitizer payloads generate reactive species (ROS) that kills cells with light and oxygen. Another example is photo-caging or photo-activation. The precise spatial and temporal control offered by photomedicine not only significantly minimizes off-site toxicities and enhances the therapeutic index, but also markedly expand the range of druggable targets.

• Effective use of generative models and machine learning to design new functional molecules
• AI-driven prediction models and optimization strategies to enhance affinity and permeability of peptide drugs and conjugates
• Need for benchmarks and optimized structure prediction algorithms​

Peptide-based therapeutics are gaining attention as they can combine the benefits of both small molecule and antibody drugs. The Gao group at Boston College contributes to peptide drug discovery by introducing new chemistries into peptide design and by developing novel display platforms. Specifically, we have been developing chemically “enhanced” phage libraries that incorporate multicyclic structures as well as designer covalent warheads.  This presentation will discuss the novel chemistries we recently developed for covalent binding of amines and for cyclizing peptides on phage. Case studies will also be presented to illustrate the scope of applications for the chemically enhanced phage libraries.

We present ring-closing amino acids bearing linkers with functional groups that enable further derivatization and efficient peptide macrocyclization. This strategy introduces a point of diversification, expanding the structural and chemical diversity of cyclic peptides. We are applying this approach to develop ligands selective for the D538G mutation in the ligand-binding domain of estrogen receptor alpha, a clinically relevant variant associated with endocrine therapy resistance. The method supports the generation of cyclic peptide libraries from a common sequence, offering a versatile platform for targeting wild-type and mutant estrogen receptors and other therapeutically important proteins.

Disease associated molecular profiles reveals new targets together with their molecular partners, interfaces, mutations, and therapeutic objectives. Some of the interfaces are not suitable for small molecules and larger oligomers, with unusual monomers and attachments needed. Constrained backbone helps to make sampling in both chemical space and conformational space feasible. We present benchmarks and optimized structure prediction algorithms to enable optimization of these binders and some success stories.

Our lab develops discrete generative models to design functional biologics directly from sequence. Our first-generation frameworks, PepMLM and PepPrCLIP, generate high-affinity peptide binders to undruggable targets, while newer design modules, like moPPIt and SOAPIA, enable motif-specific recognition and isoform-specific targeting, respectively. Our newest models, PepTune, AReUReDi, and TR2-D2 extend discrete diffusion and flow-based generation to produce noncanonical and cyclic peptides optimized for high affinity, solubility, and half-life, as well as low toxicity and non-fouling properties. We finally show that these algorithmic frameworks now generalize across modalities, supporting the design of mRNA, membrane proteins, and nanobody therapeutics.

In this talk we will discuss the development of predictive models to evaluate peptide target binding and passive diffusion across cell membranes. Application of AI-driven multi-objective optimization strategies to enhance both affinity and permeability simultaneously and case examples demonstrating how these approaches accelerate peptide drug discovery will also be highlighted.

Despite nearly a hundred years of concerted efforts, drugs comparable to antibiotics remain elusive for the treatment of cancer. One of the key impediments has been that at the cellular level cancer cells remain similar to their environment and the often subtle changes have proved hard to target and vary widely by stage and type of cancer. Many properties common to tumors have remained untargeted primarily because they are difficult to modulate via small molecules. Transformative changes in peptide technology are now paving the way to target more complex collective cellular properties. This opens up a new route towards treatment of evolving and age related diseases such as cancer and dementia, with remarkable results.

3B010, developed through our peptide discovery platform, exhibits excellent preclinical performance as a GPC3-targeting radioligand. Its high tumor retention and low kidney uptake in mouse models support its potential for clinical translation in both diagnostic imaging and radiotherapeutic applications. Human IIT studies demonstrate its high specificity and strong tumor retention in GPC3-overexpressing hepatocellular carcinoma (HCC).

Peptide drugs are potent but often short-lived. We introduce antibody–peptide fusions delivering selective GPCR agonism with potential for extended exposure. Screening leverages machine-learning–designed soluble GPCR analogs that present native-like epitopes in stable, screening-ready surrogates. These designs are structurally validated and show specific ligand binding, providing reliable tools for discovery and characterization.