Protein Stability & Formulation

Longer-acting insulin variants can be engineered by a combination of lowering the receptor affinity and extending the half-life, e.g., by fatty-diacid derivatisation for reversible albumin binding. Key to this is to avoid unwanted non-receptor-mediated clearance mechanisms. In this presentation, we demonstrate by LC-MS that insulin undergo chain-splitting into free A- and B-chains via disulfide rearrangement as an unexpected clearance mechanism, and that engineering by amino acid substitutions can reduce this.

Chugai has a clear strategy for sample preparation of forced degradation regarding to manufacturing process of each product. In addition, the samples are used for the characterisation of products including peak characterisation of purity tests and bioassay. This presentation introduces the strategy and some case studies of products, including some unique molecules developed in Chugai.

The emergence of machine learning and artificial intelligence is transforming discovery bioscience and the biologic development pipeline. In this talk, I will present two disparate methods to identify tractable protein sequences early in development—a high-throughput in vivo assay to generate large volumes of genotype-phenotype training datasets and a statistical method to identify key predictors of long term stability.

Modern antibody development hinges on accurate prediction of developability traits. However, training data are often scarce and heterogeneous. First, I will present a general ML framework to overcome this challenge, which combines pre-trained embeddings, stratified nested cross-validation, and ensemble learning to build sequence-based models that generalise to unseen regions of the sequence space. Deployed on nanobody thermostability, the resulting predictor, NanoMelt, remains accurate for sequences remote from the training set, and its forecasts accelerate nanobody selection in experimental tests. Second, we harness combinatorial droplet microfluidics to generate high-resolution solubility phase diagrams for clinical IgGs across panels of common excipients. Tens of thousands of datapoints per antibody show that, although most excipients improve solubility, the extent of that benefit is strongly molecule-dependent. Sequence- and structure-level analyses reveal the physicochemical motifs governing these responses, enabling targeted, minimal-screen formulation strategies. Together, these advances demonstrate how judicious use of limited data can unify predictive modelling and high-throughput experimentation to streamline antibody development.

Regulatory expectations, ICH Q1 revision, industry trend, applications, and challenges of the predictive stability for biologics will be discussed in this presentation.

Conducting MABEL (Minimum Anticipated Biological Effect Level) studies with ultralow concentrated clinical products presents unique challenges. These include ensuring accurate dosing, maintaining product stability, and achieving reliable bioanalytical measurements. Analytical challenges involve precise quantification at low concentrations and ensuring compatibility studies demonstrate no drug loss during administration. Specification setting is difficult due to the low concentration, requiring rigorous reasoning. 

• Methods to identify protein sequences early in development
• Bridging prediction and experimentation
• Applications in developability assessments, stability and formulation design
  

• Starting risk mitigation process early in drug development
• Incorporating advanced in silico and in vitro de-immunisaion tools into proten engineering process
• Balancing immunogenicity risk and desired biophysical properties
  

The talk will introduce our ImmTAX molecules and explain how they work. This presentation will present data on our formulation optimisation process and discuss challenges we face when delivering ImmTAX to patients.

I will share advances within application of a droplet microfluidic platform, phase scan, in biophysical characterisation, and formulation development of biologics. Data shared will be a blend of recently published and unpublished material from collaboration between the research group of Professor Tuomas Knowles at University of Cambridge and the Drug Product Research area at Novo Nordisk and include examples such as:

• Analysis of peptide precipitation 
• Analysis of mAb phase separation

And a snapshot of bayesian optimisation guided multiparameter formulation development from collaboration with Professor Paolo Arosio from ETH Zürich will be given.

Unwanted immune responses to biopharmaceuticals can affect their safety and efficacy. Where necessary, unintended immunogenicity will be mitigated in the clinic, but mitigation should preferably start early in development, at the drug design phase. This presentation discusses incorporating advanced in silico and in vitro de-immunisation tools into protein engineering processes to select a lead candidate that balances immunogenicity risk and desired biophysical properties.

The ability of proteins to aggregate is a major challenge upon the development of therapeutic products. Understanding how protein-protein and protein-solvent interactions determine the stability, self-assembly kinetics and morphology of the aggregates is a conditio sine qua non to unravel the mechanisms ruling the self-assembly reaction. I will present our approach based on fluorescence microscopy, X-ray scattering and spectroscopy, aimed at identifying the interactions responsible for protein and antibody phase separation, conformational changes and colloidal instability, and how those aspects are linked to the variability in the aggregate morphologies. I will also discuss the potential role of different aggregates in inducing immunoresponse depending on their structure, size, and morphology. Our findings provide a scenario in which heterogeneous structures can be generated as a result of interconnected aggregation pathways, being this aspect of relevance for the design of stable pharmaceutical formulations.