Macrocyclic peptides are valuable molecular formats for drug development, bridging small molecules and larger biologics. We recently developed a yeast-based combinatorial platform for the quantitative screening and characterisation of these ligands. By integrating flow cytometry and next-generation sequencing, the platform screens millions of variants hourly, enabling real-time monitoring and tuning of binding properties. We will present our progress in evolving macrocyclic peptide ligands against diverse protein targets.

Discovering binders and modulating interactions are two of the central challenges in chemical biology and drug discovery. Here I describe two complementary platforms developed in our lab. PANCS-Binders enables rapid, scalable discovery of miniprotein and peptide-based binders to targets of against. PANCS-glues leverages intrinsically disordered peptides to drive cooperative, context-independent protein-protein interactions. Together, these technologies provide powerful new strategies for discovering molecules that modulate biological systems.

Tumour-associated glycans Tn and STn are attractive but underexploited oncology targets due to challenges in achieving selective targeting. We present a structure-guided phage display platform generating monoclonal antibodies against glycopeptide epitopes (“combotopes”) by combining glycan-binding VH domains with peptide-selective VL diversity. The platform enables both discovery of new high-affinity antibodies and rapid optimisation of tumour-selective specificity for therapeutic development.

This presentation describes the application of Deep Screening and Machine Learning to augment AstraZeneca’s biologics discovery pipeline. This unique combination of cutting-edge technologies expedites the delivery of candidate biologic molecules with the desired physical and functional characteristics needed to make potent, efficacious and developable drugs to treat patients with clinical unmet needs.

This presentation highlights how next-generation sequencing (NGS) of VHH immune repertoires after immunization and display-based screening can be combined with data-driven prioritization and AI-enabled sequence design to support single domain antibody (sdAb) hit discovery. Our workflow includes a One Shot humanization and multi-parameter optimization approach using language models to generate humanized, developability-aware candidates early in the process. Scope, practical limitations, and hit rates will be discussed.

Antibody display technologies coupled to next-generation sequencing (NGS) techniques have been revolutionizing monoclonal antibody (mAb) discovery. However, with deep sequencing capabilities arises the difficulty of producing and testing all the selectable candidates from the big NGS datasets. iSeRa-AI tool is thus being built to better select candidates, extract more information from the datasets and automate the process. iSeRa-AI selects candidates by analysing the entire dataset and taking into account similarity between sequences, counter-selection, traits of non-specificity and of bad developability. It goes even further by suggesting artificial candidates (i.e., not present in the dataset) to be likely good binders. Validation is ongoing both with in house generated and publicly available datasets. Here, the latest developments and performance of the new iSeRa-AI tool will be presented.