Using features calculated from intrinsic sequences and molecular dynamics simulations of 20 full-length IgGs, we have applied machine learning to develop a k-nearest neighbors regression model with two features – spatial positive charge map (CDRH2) and solvent accessible surface area of hydrophobic residues – to predict aggregation rates and a logistic regression classification model with two features – spatial negative charge map (HC-Fv) and spatial negative charge map (LC-Fv) to predict viscosity.

Poor developability parameters are among the major causes of attrition in antibody discovery programs, in part because it is challenging to assess them experimentally in a high-throughput manner. The availability of large databases of antibody sequences and developability measurements is making it possible to develop computational methods to predict these parameters from their amino acid sequences. I will present methods to achieve this goal.

Our teams have utilized innovative computational protein design algorithms and protein subunit-based screening approaches to generate bifunctional biologics with antibody-like biophysical properties. More recently, we have pushed these boundaries to generate multifunctional biologics with the goal of achieving highly specific cellular targeting of various functionalities. Future directions of multifunctional biologics will be discussed, including the benefits of protein and antibody subunit design, tailored computational methodologies to solve specific design parameters, and strengths and weaknesses of the approaches.

T cell-engaging antibodies represent a promising class of bispecifics in immuno-oncology. Here we leverage the two well-characterized modalities within Pfizer, diabody-Fc, and full-length IgG2, and a series of engineered cell lines, to interrogate how we might use bispecific antibodies to modulate T-cell activation while using T cell-redirected cytotoxicity and cytokine release as two primary readouts. We demonstrated the possibility of decoupling cytotoxicity and cytokine release if one optimizes inter-membrane distance via a combination of epitope to target and modality. 

Bispecific antibodies (BsAbs) often exhibit a broad range of pharmacokinetics (PK). Optimization of the neonatal Fc receptor (FcRn) interactions and removal of undesirable physiochemical properties have been used to improve the ‘pharmacokinetic developability’ for mAb therapeutics, yet there is a sparsity of such information for BsAbs. The present work evaluated the influence of FcRn interactions and inherent physiochemical properties on the PK BsAb formats with different fusion partners and configurations.  

T cell redirecting antibodies have gained momentum in cancer immunotherapy with limitations in the clinic due to toxicities associated with CRS. We developed AMG 340, an anti-CD3xPSMA for the treatment of mCRPC that exhibited killing of PSMA+ tumor cells comparable to a positive control with the same anti-PSMA but stronger anti-CD3 arm, but significantly reduced cytokine secretion compared to the positive control, potentially having a better clinical outcome for patients. AMG 340 has favorable developability characteristics and pK profile, allowing Q3W dosing, and is currently in Phase 1 clinical trials.

There are many antibody discovery and optimization technologies available today. Each has their advantages and limitations. At Mosaic, we utilize technologies that are fit for purpose for any drug discovery challenge we encounter and are agnostic as to whether it is a technology we provide ourselves or work with partners to deliver. Herein, we will describe three drug discovery challenges that were solved using multiple approaches. 

Immunogenic responses, such as generation of anti-drug antibodies (ADA), against biotherapeutic products may have detrimental effects on drug safety, efficacy, and pharmacokinetics. Clinical trials are aimed at directly examining the ADA response in patients, but what if we could predict immunogenicity before clinical trials begin and mitigate those immunogenicity risks? This presentation will focus on in silico and in vitro tools to assess immunogenicity risk of monospecific and bispecific biotherapeutics.

• The biparatopic format of zanidatamab zovodotin enables enhanced tumour cell binding, internalization, and payload delivery compared to a monospecific HER2-targeting ADC. 
• Zanidatamab zovodotin induces in vitro hallmarks of immunogenic cell death (ICD) including increased calreticulin exposure, increased high mobility group box-1 (HMGB1) exposure, and increased extracellular ATP secretion in a HER2-dependent manner. 
• Taken together, the differentiated properties of zanidatamab zovodotin compared to other HER2+ targeted therapies, supported progression to the clinic where durable responses have been observed in heavily pretreated HER2+ patients.?

FRa is a clinically-validated ADC target. The scientific rationale of a next-generation FRa-specific IMGN151, its design, and preclinical activity will be discussed.