As a prime next-generation therapeutic target for obesity and metabolic disorders, GIPR belongs to the GPCR transmembrane protein family. It features unstable conformation, high screening difficulty, and inherent trade-offs between candidate molecular affinity and developability, making it a widely recognized challenging target for drug development in the industry. Developing a qualified anti-GIPR antibody via conventional technical routes typically takes 18–24 months and is plagued by three core industry pain points:

• Difficulties meeting affinity and cellular function benchmarks: Antibodies derived from traditional immunization and library screening exhibit weak binding capacity and fail to match 2G10, the parent antibody of the leading overseas candidate AMG133;

• Prolonged screening cycles: Animal immunization plus multiple rounds of in vitro and in vivo optimization and verification result in a full development cycle generally exceeding 18 months;

• Late exposure of developability defects: CMC challenges including non-specific binding, high immunogenicity and propensity for aggregation are only uncovered after molecular screening is completed, rendering substantial upfront R&D investments futile.

 

Great Bay Bio selected GIPR antibody design as a case study to validate the capabilities of AlfaBodY, its proprietary AI-powered intelligent molecular design platform. The full R&D workflow was completed within merely 8 months, with candidate molecules rapidly generated and validated via in vitro and in vivo efficacy assays. This fully demonstrates the robust capability of the AI platform to tackle challenging GPCR targets and rapidly generate differentiated antibodies.

 

Relevant platform research titled Generation of High-Affinity Anti-GIPR Antagonist Antibodies with Sustained and Non-rebound Weight Loss in DIO Mice by AlfaBodY has been published on bioRxiv, the international preprint platform for biomedical research. The paper comprehensively elaborates the standardized end-to-end workflow of AlfaBodY, covering structural prediction, antibody sequence generation and virtual screening, multi-layered in vitro screening, and in vivo efficacy verification in obese animal models. This complete GIPR antibody case verifies that the AlfaBodY AI antibody design platform can efficiently conquer traditionally intractable GPCR targets and deliver a replicable, high-efficiency new paradigm for novel antibody drug development.

 

▲ Research outcomes of the AlfaBodY platform are available on bioRxiv

 

Platform Case: High-Performance Validated GIPR Molecules Generated in 8 Months

The AlfaBodY platform drives the entire development workflow of anti-GIPR antagonist monoclonal antibodies, accelerating the closed-loop process spanning structural prediction, antibody design, multi-dimensional in vitro validation, developability assessment and in vivo efficacy evaluation.

 

The platform completed a full spectrum of tests including BLI/SPR affinity measurement, cellular binding assays, cAMP functional tests, non-specific binding assessment, developability evaluation, and in vivo efficacy studies in DIO obese mice. Successful development of candidate molecules achieved an effective breakthrough against this hard-to-drug GPCR target. Within 8 months, the platform accomplished work that conventionally requires 18–24 months for full GIPR antibody development.

 

▲ Left and right panels show predicted binding structures of candidate antibody molecules AB106-156 and AB106-131 bound to the antigen, respectively

 

AI-Directed Molecular Design Achieves Nearly 10-Fold Affinity Improvement

Leveraging atomic-level antigen-antibody binding simulation algorithms, AlfaBodY generates antibody sequences targeting the extracellular domain of GIPR. Through four rounds of AI antibody sequence generation, a total of 157 antibody molecules were produced. High-affinity candidates were identified via wet-lab screening, eliminating the need for massive random clone screening and enabling rapid identification of potent candidates.

 

▲ The figure illustrates the binding affinity ratio of antibodies generated in each design round relative to the reference antibody 2G10. Binding affinity was markedly enhanced following four rounds of optimization.

 

Affinity measurements via bio-layer interferometry (BLI) show that AB106-131 exhibits an affinity of 1.2 nM and AB106-156 reaches 1.7 nM, representing a 10-fold and 7-fold improvement over 2G10 respectively. Flow cytometry was used to measure antibody binding to HEK293-T cells stably expressing GIPR (P48546). The EC₅₀ values of AB106-156, AB106-131 and 2G10 were 3.2 nM, 2.2 nM and 2.8 nM correspondingly. A cAMP signaling reporter cell line constructed using GIPR-overexpressing cells was adopted to evaluate antibody-mediated blockade of the GIP/GIPR signaling pathway. The IC₅₀ values of AB106-156, AB106-131 and 2G10 were 5.9 nM, 4.4 nM and 4.0 nM. The two candidate molecules display cellular binding and inhibitory function comparable to 2G10.

 

AI Explores Broader Sequence Space

Sequence identity between the AI-generated antibody sequences AB106-131 / AB106-156 and the parent antibody stands at 74.4% and 81.9%, while CDR region identity reaches 71.7% and 75.0%. This confirms that AI can substantially expand sequence diversity while boosting biological activity.

 

Pre-emptive Off-Target Risk Screening to Ensure Safety

AlfaBodY integrates a built-in off-target binding prediction model to eliminate sequences prone to non-specific binding at the molecular design stage. Subsequent FACS cellular assays corroborated the platform’s predictive results: in tests on GIPR-negative HEK293T cells, the fluorescence MFI values of the AB106 series matched those of blank PBS and benchmark control antibodies, indicating zero risk of off-target binding. Developability is guaranteed from the outset, avoiding repeated optimization and correction in later stages.

 

Synergy with AlfaDAX for Developability Prediction to Mitigate CMC Risks

Combined with AlfaDAX, the proprietary developability prediction system, AlfaBodY evaluates and optimizes more than ten key parameters synchronously during molecular design, including surface charge, aggregation propensity, viscosity, non-specific binding and immunogenicity.

 

AB106-156 achieves an immunogenicity score of only 0.13, significantly outperforming the benchmark molecule. It satisfies all criteria for molecular stability, freeze-thaw tolerance and accelerated stability under 40°C, featuring low aggregation propensity and compatibility with high-concentration formulations. Supported by upfront platform evaluation, the project simultaneously advanced CHO cell line construction, culture medium optimization and early production process development.

Successful validation of this workflow proves that the platform enables developability and CMC considerations to be embedded at the design stage, helping partners prevent project rework triggered by developability issues in later phases.

 

• Isoelectric point (pI): Proteins are the most unstable and have the lowest solubility at their pI due to the absence of charge repulsion effects, making them prone to aggregation and precipitation.

• Humanization score (Hu): The higher the humanization score, the greater the degree of humanization. A score < 0.2 indicates a non-humanized ant•ibody, a score ranges from 0.2 to 0.6 indicates a humanized antibody, and a score > 0.6 indicates a fully humanized antibody.

• Stability (Stab): A score > -40 indicates a risk of low molecule stability.

• Aggregation and precipitation (AP): A score > 1 indicates a risk of aggregation and precipitation.

• Viscosity score (Vis): A score > 1 suggests a risk of excessive viscosity at high concentration (> 150 mg/ml).

• Non-specific binding (NSB): A score > 1 indicates a tendency towards non-specific binding.

• Antibody-antigen predicted template modelling score (abPTM): A score < 0.6 indicates that the predicted structure is not reliable. The higher the abPTM, the higher accuracy in the antigen-antibody docking prediction.

• Blocking rate (block_rate): The blocking rate describes the overlap exists between the epitopes of a ligand and an antibody on the same antigen.

• Binding energy (Energy): A binding energy > -5 suggests a high binding energy between the target and the predicted antibody structure, indicating potential inaccuracy in the predicted antibody structure.

• Immunogenicity score (Immu): A score > 1 indicates a risk of high immunogenicity (ADA >5%).

 

 

In Vivo Efficacy Validates the Strength of the AI Platform

AB106-156 designed via AlfaBodY demonstrates core advantages absent from current mainstream weight-loss therapeutics when tested in the hGIPR DIO obese mouse model:

1. Potent weight reduction and synergistic efficacy in combination: Monotherapy with AB106-156 delivered a maximum weight loss of 18.2%, superior to semaglutide’s 15.5%. Combined administration achieved peak weight loss of 25.4%, showing remarkable synergistic effects via dual-pathway modulation.

2. No weight rebound after treatment cessation for sustained metabolic stability: During the 19-day recovery phase following drug withdrawal, the semaglutide group exhibited a notable weight rebound of 10.7%, whereas the AI-generated AB106-156 only rebounded by 2.7%, with continuous improvement in glycemic control after treatment stop.

3. Selective fat loss with preserved lean mass: MRI body composition analysis revealed that weight loss induced by semaglutide was accompanied by lean mass loss and rapid fat regain after drug withdrawal. AB106-156 selectively consumes adipose tissue with nearly no loss of lean mass, enabling long-term improvement of body composition.

 

Horizontal Benchmarking of Same-Target Programs

A conventional CRO screened over 2,000 molecules via protein and cellular immunization. Although molecules with affinity close to the benchmark were obtained, their cellular function remained inferior to the benchmark even after affinity maturation; some exhibited no blocking activity at all.

 

In contrast to the two-year development cycle of traditional technologies, the AlfaBodY platform generated and validated high-quality GIPR molecules within 8 months. This complete development case intuitively illustrates the disruptive empowering effect of AI on original innovative drug discovery, highlighting the platform’s strengths in accelerating progress, improving quality and enabling differentiated innovation.

 

As a leading provider of AI macromolecular design platforms in the industry, Great Bay Bio will continue to leverage the AlfaBodY intelligent molecular design platform. Through diversified collaboration models including platform licensing, co-development and technical services, we aim to support global pharmaceutical enterprises and research institutions in tackling challenging targets such as various GPCRs and transmembrane proteins, and accelerate the delivery of more globally competitive innovative medicines.