Therapeutic Antibody Landscape
Over 200 approved therapeutic antibodies are commercially available, with significant growth in the last 15 years. Despite their popularity there are various pain points associated with traditional antibody therapeutics including high costs, long development timelines and limited search regions.
GenScript’s AI-Powered Antibody Solutions
Robert Ford, Field Application Scientist at GenScript, introduced some of the cutting-edge solutions the company offers scientists working with antibodies to help them advance their pipelines. Like many companies GenScript is leveraging AI to improve their services on offer and help scientists speed up their workflows.
GenScript’s immunoinformatic platform relies on an epitope prediction model as well as a codon optimization tool. The platform also offers various developability assessments such as immunogenicity prediction, viscosity, aggregation, degradation and structure annotation.
Ford explained that GenScript has developed AI tools to assist with antibody engineering, humanisation as well as pet care with a specific focus on binding affinity maturation. Now, GenScript is working on multi-species humanisation tools, LLM-based antibody binding affinity maturation, and multi-modal evolution-based therapeutic developability models.
Epitope Prediction and Codon Optimisation Tools
GenScript’s epitope prediction AI model was compared with EpiDope’s antigen prediction model across several antigens. The results showed that GenScript’s epitope prediction model performs better than EpiDope’s, with more true negatives and fewer false positives in antigen selection.
Once antigen selection is complete, one may want to express it recombinantly, this is where GenScript’s codon optimisation tool, GenSmart ZQ comes in handy. Ford stated that this tool has been tried and tested against competitors and the GenSmart ZQ came out on top.
Developability and Structural Annotation Tools
Regarding developability assessment, GenScript uses Mab chain domain annotator and a bispecific structure annotation tool. Researchers can input their sequence, and the tool colour codes the exact regions where the Fc, Fv and other linkers are located. These tools can also recognise and distinguish between a human, mouse or rabbit antibody.
For the bispecific structure annotation, one can input an antibody sequence and if the platform can identify it in the database, it can give the client the sequence mode, class and subclass of the antibody. In cases where the antibody is not in the GenScript database it will give a relative sequence mode.
Degradation, Viscosity, and Immunogenicity Prediction
GenScript’s degradation site prediction tool mainly focuses on asparagine deamidation and methionine oxidation. This technology has a high accuracy level of 94%. Furthermore, the viscosity prediction and immunogenicity prediction models are based on LLM models. The ProtT5 (GenScript’s viscosity model) and GenScript's immunogenicity model outperformed other models in the market offered by Pfizer and the University of Michigan.
Bispecific Antibodies: Function and Challenges
The second part of the workshop delved into the mechanisms of action for bispecific antibodies and bispecific antibody production. Bispecific antibodies have two distinct binding epitopes, meaning one can either bind the bispecific antibody to two different antigens of the same protein or two separate epitopes of the same antigen. This makes them more effective in tackling complex diseases compared to monoclonal antibodies.
The majority of bispecific antibodies are T cell recruiters, binding CD3 and tumour-associated antigens to bring T cells into close proximity with target cells, resulting in cytotoxicity. Other mechanisms include receptor inhibition, activation, and cofactor mimetic actions.
Ford added that producing bispecific antibodies is challenging due to the need for correct heavy and light chain formation. Techniques like knob-into-holes technology and CrossMAb technology help overcome these challenges by ensuring proper chain pairing.
TurboCHO Platform and Case Studies
Ford introduced the TurboCHO platform and CrossMAb technology and then gave some case studies from various bispecific antibody formats. The TurboCHO platform enhances bispecific antibody production through iterative optimisation of cell engineering, medium, transfection reagents, and production processes. Ford highlighted case studies that demonstrated improved yield and purity of bispecific antibodies using this platform.
The first case study looked at a Fab scFv-Fc bispecific format and expressed it using an unoptimised system with Expi-293F and compared this to GenScript’s TurboCHO platform. The Expi 293F platform performed poorly, giving a purity of around 40% and a yield of just 16 mg/l but with some tweaks, Ford was able to significantly increase the yield and quality of the final product. However, the TurboCHO platform performed significantly better, giving 90% purity and a final product of 340mg/l.
The second case study examined the use of CrossMAb and Glofitamab, a T cell engager. Ford mentioned that the VH-VL charge interactions avoid Bence-Jones-like side products. Moreover The chain of the molecule underwent ZQ optimisation and two purification steps.
Collaborative Research with Liverpool School of Tropical Medicine
From a collaborative perspective, GenScript is working with the Liverpool School of Tropical Medicine on hijacking the hemagglutinin-sialic acid interaction to prevent viral entry. The collaboration with GenScript has expedited the synthesis and characterization of FC fragments for this purpose. A team at the Liverpool School of Tropical Medicine aims to test FC fragments on live influenza viruses and present further data at Immunology 2025 in Hawaii.