Zero-Shot Design of a Biobetter Cetuximab: Enhancing EGFR Binding with ConvergeAB™

TL;DR ConvergeAB™ was used in a zero-shot setting to design a biobetter version of cetuximab targeting EGFR. The lead Converge-designed antibody demonstrated consistently superior binding kinetics, with ~2.1× higher affinity than cetuximab and ~4.4× higher than a computationally designed competitor.

Introduction

Epidermal Growth Factor Receptor (EGFR) is a transmembrane tyrosine kinase that plays a central role in regulating cell proliferation, survival, and differentiation. Upon binding of its natural ligands, such as epidermal growth factor (EGF), EGFR undergoes dimerization and autophosphorylation, triggering downstream signaling cascades. Dysregulation of EGFR, through overexpression or mutation, is a well-established driver of oncogenesis. Elevated EGFR activity has been implicated across multiple solid tumors, most notably in colorectal cancer, head and neck squamous cell carcinoma (HNSCC), and non-small cell lung cancer (NSCLC). In these contexts, aberrant EGFR signaling promotes uncontrolled cell growth, resistance to apoptosis, and enhanced metastatic potential, making it a compelling therapeutic target.

Cetuximab is a chimeric monoclonal antibody designed to target the extracellular domain of EGFR. By binding to the receptor with high specificity, cetuximab sterically blocks ligand interaction, thereby preventing receptor activation and downstream signaling. Clinically, cetuximab has become a cornerstone therapy in EGFR-driven malignancies, particularly in metastatic colorectal cancer and recurrent or metastatic HNSCC. It is widely used in combination with chemotherapy or radiation, with global adoption supported by regulatory approvals across the world.

At the molecular level, the therapeutic efficacy of cetuximab is fundamentally governed by its competition with endogenous ligands such as EGF for binding to EGFR. This competitive interaction is dictated by binding affinity, commonly quantified by the dissociation constant (KD). A lower KD reflects tighter binding and greater competitive advantage. In tumors where ligand concentrations are high, cetuximab must effectively outcompete EGF to suppress receptor activation. Insufficient binding affinity can allow residual signaling, contributing to suboptimal response and disease progression. Therefore, optimizing KD is not merely a biochemical consideration but a critical determinant of clinical efficacy, influencing both depth and durability of response in patients.

Figure 1. Competitive binding at EGFR is a key determinant of the therapeutic efficacy of anti-EGFR antibodies. 

Epidermal growth factor (EGF, red) and anti-EGFR monoclonal IgG (green) compete for binding to the EGFR transmembrane receptor on the surface of cancer cells. Engagement of EGFR by EGF activates proliferative signaling pathways that drive tumor growth and progression. In contrast, antibody binding blocks ligand-induced receptor activation, leading to growth inhibition, cell cycle arrest, or apoptosis. The outcome of this competition is governed by binding kinetics, with lower KD antibodies exhibiting superior competitive advantage over endogenous ligands.

Results

To generate a biobetter version of cetuximab, we prompted ConvergeAB™ with the original cetuximab sequence together with the EGFR target sequence. No task-specific or ad hoc training was done prior to the initiation of this affinity maturation campaign, effectively creating a zero shot effort. In brief, ConvergeAB™ is a multi-step platform in which sequence diversity is introduced into a seed antibody using a proprietary protein language model, generating ~100,000 target-aware candidates. These candidates are subsequently filtered through a series of explicit and orthogonal predictors that evaluate key properties, including binding affinity, structural compatibility (docking), expressibility, thermal stability, and solubility. Additional details on the ConvergeAB™ platform can be found here: https://converge-bio.com/solutions/converge-ab.

The top 10 predicted candidates of the campaign were selected for experimental validation. Each variant was expressed and screened for binding using single-concentration surface plasmon resonance (SPR). The highest-performing candidate from this screen was advanced for full kinetic characterization, including precise KD determination. As benchmarks, we included both the original cetuximab and a recently published anti-EGFR antibody from Cradle Bio, reported as a top performer in a protein design competition (https://www.adaptyvbio.com/blog/po104/).

All three antibodies (Converge, Cradle, and cetuximab) were produced in the IgG1 format, consistent with the clinical form of cetuximab. For rigorous head-to-head comparison, the three antibodies were analyzed side-by-side in two unique SPR repeats (#1 and #2) using a six-analyte, multi-cycle kinetic SPR assay. Across all measured parameters (kon, koff, and KD), and in both biological repeats, the Converge-designed antibody consistently demonstrated superior performance, achieving an average binding affinity 2.1-fold stronger than cetuximab and 4.4-fold stronger than the Cradle-derived antibody.

Table 1. Summary of SPR kinetic parameters for anti-EGFR antibodies. Three anti-EGFR antibodies were evaluated side-by-side for binding to EGFR. Each SPR repeat represents an independent, parallel measurement of all three antibodies under identical conditions. The association rate constant (kon) reflects the rate of antibody–antigen complex formation, with higher values indicating faster binding. The dissociation rate constant (koff) reflects the rate of complex dissociation, with lower values indicating greater complex stability. The equilibrium dissociation constant (KD), calculated as koff / kon, summarizes overall binding affinity, with lower values corresponding to stronger binding. 

Figure 2. SPR multikinetic raw data of both repeats. SPR sensorgrams from each repeat were normalized to enable side-by-side comparison. EGFR was used as the analyte at the following concentrations (nM): 0.2, 0.39, 0.78, 1.6, 3.1, 6.2, and 12.5. Curve thickness corresponds to analyte concentration, with the boldest curve representing the highest concentration.

Sequence analysis reveals the Converge sequence had a total of six edits compared to cetuximab, comprising four substitutions to alanine (VH T61A, VH S87A, VL V9A, and VL N93A) and two asparagine-to-aspartate substitutions (VH N88D and VL N32D), distributed across both framework and CDR regions. To assess whether these affinity gains were associated with large-scale structural remodeling, we predicted the 3D structures of both antibodies and aligned them. The resulting overall RMSD was very low, at 0.16 Å, indicating that the Converge variant preserves the global cetuximab binding architecture. This is an important observation, as it argues against a wholesale change in fold or binding mode and instead supports a mechanism in which affinity improvement is driven by highly localized structural and energetic optimization within an otherwise conserved paratope.

Figure 3. Sequence alignment of cetuximab and Converge-designed anti-EGFR antibody. Amino acid sequence alignment comparing cetuximab with the Converge-derived anti-EGFR IgG marks the positions of the six edits introduced relative to cetuximab.

This interpretation is consistent with the nature of the edits. The alanine substitutions are expected to reduce side-chain bulk and remove potentially suboptimal polar functionality, which can improve local packing, decrease conformational heterogeneity, and favor a more pre-organized binding-competent state. Likewise, the Aspargene to Aspartate substitutions introduce negative charge and altered hydrogen-bonding geometry, which may strengthen electrostatic complementarity with EGFR or stabilize productive local conformations. The structural comparison further supports this localized mechanism: although the global RMSD remains minimal, the N93A substitution produces a 0.66 Å shift at the Ca position and displaces Trp94 by 0.5 Å. These are small but meaningful local rearrangements, especially in an antibody paratope, where sub-angstrom movements can measurably alter side-chain presentation, loop packing, and epitope complementarity without changing the overall fold.

Figure 4. Structural alignment of cetuximab and Converge-designed anti-EGFR antibody. Structural superposition of cetuximab and the Converge-derived anti-EGFR IgG highlights overall architectural similarity.

Taken together, the sequence and structure analyses support a compelling model for the experimentally confirmed affinity improvement. Rather than arising from major structural divergence from cetuximab, the Converge sequence appears to retain the parental architecture while introducing a set of cooperative micro-adjustments that fine-tune the binding interface. The very low overall RMSD, combined with specific local shifts associated with N109A and neighboring Trp114, is therefore fully consistent with an affinity maturation mechanism based on paratope preorganization, improved local packing, and targeted electrostatic optimization.

Conclusion

In this work, we demonstrate that ConvergeAB™ can be applied in a fully zero-shot setting to generate a biobetter antibody with meaningfully improved properties over a clinically established therapeutic. Starting from cetuximab, the platform produced candidates that translated from in silico predictions to experimentally validated gains in binding kinetics, ultimately yielding an antibody with substantially enhanced affinity to EGFR. These results highlight the ability of ConvergeAB™ to efficiently navigate the antibody design space, and its potential to accelerate the development of next-generation biologics with improved therapeutic performance.