Bispecific Antibody Engineering

A heterodimerization strategy that forces correct heavy-chain pairing through steric complementarity. A bulky 'knob' residue (e.g., T366W) is introduced into the CH3 domain of one heavy chain, while a corresponding 'hole' (e.g., T366S/L368A/Y407V) is created in the other.

• Computational challenge: Predicting the stability of the engineered CH3 interface using Rosetta or FoldX
• Pairing efficiency: >95% heterodimer purity when optimized
• Origin: Developed by Genentech; foundational to many clinical-stage bispecifics

A domain-swapping approach that resolves the light-chain mispairing problem by exchanging entire Fab domains between the two antibody arms. In the CrossMabFab variant, the CL and CH1 domains are swapped on one arm, ensuring each light chain pairs only with its cognate heavy chain.

• Key advantage: Maintains native IgG architecture without introducing artificial linkers
• Computational need: Molecular dynamics to verify that domain swapping does not introduce structural strain or aggregation propensity
• Variants: CrossMabFab, CrossMabVH-VL, CrossMabCH1-CL

A minimalist format consisting of two single-chain variable fragments (scFvs) connected by a flexible glycine-serine linker. One scFv targets a tumor-associated antigen, while the other binds CD3ε on T cells, creating an artificial immunological synapse.

• Molecular weight: ~55 kDa, enabling rapid tumor penetration but short serum half-life (~2 hours)
• Computational design: Linker length optimization to prevent inter-scFv domain swapping and aggregation
• Example: Blinatumomab (Blincyto) for B-cell acute lymphoblastic leukemia

A format where the VL and VH domains of a second antibody are fused via flexible linkers to the N-termini of the light and heavy chains of a primary antibody, creating a tetravalent molecule with dual specificity.

• Architecture: Outer Fv binds antigen A; inner Fv binds antigen B
• Computational challenge: Modeling steric accessibility of the inner Fv domain, which can be occluded by the outer domain
• Linker engineering: Critical for maintaining independent antigen binding without steric hindrance

A rational design approach that introduces complementary charged residues at the CH3-CH3 interface to favor heterodimerization over homodimerization. Positive charges (e.g., K) on one chain pair with negative charges (e.g., D or E) on the other, while like-charge repulsion disfavors homodimers.

• Computational tooling: Poisson-Boltzmann electrostatic calculations to identify optimal mutation sites
• Advantage: Avoids introducing non-native steric bulk that could affect stability
• Combination strategy: Often paired with KiH for orthogonal pairing assurance

A genetic engineering approach that uses a single identical light chain for both Fab arms, eliminating light-chain mispairing entirely. The bispecificity is encoded solely in the two distinct heavy chains.

• Discovery method: Transgenic mice (e.g., VelocImmune) or phage display libraries screened with a fixed light chain
• Computational task: Identifying a light chain that is promiscuous enough to pair productively with two different heavy chains while maintaining affinity
• Advantage: Simplifies manufacturing to a single light chain plus two heavy chains

A pioneering heterodimerization strategy that forces correct heavy-chain pairing by engineering sterically complementary mutations in the CH3 domain. A 'knob' mutation (e.g., T366W) is introduced on one heavy chain, while a 'hole' (e.g., T366S/L368A/Y407V) is created on the other. Computational modeling is essential to predict the structural consequences of these mutations on thermal stability and to ensure the engineered interface does not create new immunogenic epitopes. This approach prevents the formation of non-functional homodimers, which are a major contaminant in bispecific production.

A design approach that circumvents the light-chain mispairing problem by using an identical light chain for both antigen-binding arms. This requires extensive computational screening to identify a single light chain that is functionally compatible with two distinct heavy chains without compromising the affinity of either paratope. Transgenic platforms (e.g., common light chain mice) are often used to generate the initial repertoire, but AI-driven sequence optimization is critical for refining the shared light chain to achieve high-affinity, dual-specific binding.

A domain-swapping technique that resolves the light chain mispairing issue without requiring a common light chain. In the CrossMab format, the CL and CH1 domains are exchanged in one of the Fab arms, ensuring that each light chain can only pair with its cognate heavy chain. Computational structural biology validates that the domain crossover does not introduce strain or alter the relative orientation of the variable domains, preserving paratope geometry and antigen-binding function.

A class of bispecifics that simultaneously bind CD3 on T-cells and a tumor-associated antigen (TAA) on cancer cells, forcing the formation of an immunological synapse. Unlike full-length IgG-like bispecifics, BiTEs are composed of two tandem single-chain variable fragments (scFvs) linked by a flexible glycine-serine linker. Computational modeling of linker length and composition is critical to prevent interdomain steric hindrance and to optimize the distance required for effective T-cell activation and target cell lysis.

A computational framework that simultaneously optimizes a bispecific antibody for multiple, often conflicting, properties. The algorithm must balance dual-antigen affinity, heterodimerization specificity, thermal stability, and low immunogenicity. This is typically achieved using Bayesian optimization or evolutionary algorithms that navigate a Pareto frontier of design trade-offs. The goal is to identify a sequence that does not sacrifice one critical property—such as solubility—to achieve marginal gains in another, like binding kinetics.

A family of bispecific antibodies that retain the Fc region of a conventional IgG, conferring a long circulatory half-life through FcRn-mediated recycling and the potential for effector functions like ADCC. These formats require sophisticated engineering to solve the heavy-chain heterodimerization problem. Examples include the DuoBody platform (controlled Fab-arm exchange) and SEEDbody technology (alternating IgA/IgG CH3 sequences). Computational tools predict the thermodynamic favorability of heterodimer formation over homodimer byproducts.