Humanization calculation method and structural feature analysis of nanobodies！-Antibody Detection Service Topic-Tekbiotech-Yeast and Phage Display CRO, Expert in Nano-body and Antibody Drug Development

Humanization calculation method and structural feature analysis of nanobodies！

To become an effective therapeutic, an antibody must be tolerated by the human immune system. To minimize the risk of unintended immunogenicity, a rational strategy is to maximize the "human-ness" of a drug candidate. However, although new technologies continue to emerge — many of them derived directly from fully human gene segments — most antibody therapeutics are still of non-human origin and are humanized to improve their compatibility with the human body. Early experimental humanization strategies focused mainly on grafting the complementarity-determining region (CDR) loops onto human frameworks. This approach was instrumental in establishing the dominance of this development route, but it did not take into account the specific context of each antibody sequence, which compromised its success rate. In recent years, considerable effort has been devoted to developing computational methods capable of addressing these issues. This article focuses on the structural features of nanobodies and on computational methods for their humanization.

Figure 1 Traditional IgG humanization strategies (chimerization, CDR grafting, SDR grafting and resurfacing) ^[1]

一、Structural Features of Nanobodies:

Derived from camelids (VHHs) or cartilaginous fish (VNARs), nanobodies are emerging as a promising therapeutic modality. Their small size facilitates expression, improves tumor penetration and enhances solubility, while maintaining binding affinities comparable to those of conventional antibodies. However, the structural differences between conventional antibodies and nanobodies affect the way in which they interact with antigens. Consequently, many humanization strategies — particularly computational tools — designed for conventional antibodies may not be directly applicable to nanobodies.

Figure 2 Structural features of IgG antibodies and nanobodies ^[1]

二、Applicability of Antibody Humanization Software to Nanobodies:

Given the increasingly broad therapeutic application of nanobodies, the applicability of humanization methods to nanobodies should be taken into account in future method development. The general principles of experimental nanobody humanization are similar to those for conventional antibodies, with CDR grafting, back mutation and resurfacing, for example, serving as templates. However, additional considerations are required because conventional antibodies and nanobodies have different structural features. Vincke et al.^[2] described the humanization of camelid VHHs and highlighted the importance of framework region 2 (FR2) residues in influencing binding affinity. These hallmark residues are located at positions corresponding to the VH–VL interface of conventional antibodies, are generally hydrophilic, and contribute to the improved solubility of single-domain antibodies. Because these residues occupy a buried, hydrophobic position in conventional antibodies, computational tools designed for conventional antibodies may recommend mutating them when constructing humanized models. Although this may enhance human-ness, it may also compromise their advantageous properties.

三、Dedicated Computational Methods for Nanobody Humanization:

Computational methods that take the nanobody-specific topology into account: NanoBodyBuilder2, a structure prediction tool that outperforms AlphaFold2 by 0.55 ? in the CDR3 region, is a prime example of the advantage offered by nanobody-specific tools. Software development for nanobody humanization is still at a relatively early stage, and the first such tool was Llamanade. By comparison with IgGs, Sang et al. [3] identified nanobody-specific properties and used them as the basis for rational humanization design, avoiding the humanization of residues that are critical to the physicochemical properties of nanobodies — for example, the highly conserved framework residues in the FR2 region and residues that may affect the conformation of the CDR3 loop. AbNatiV likewise balances nativeness (retention of the key residues underlying the unique structural properties of nanobodies) against humanization; it is a deep-learning-based humanization pipeline for nanobodies (and conventional antibodies). It trains a vector-quantized variational autoencoder (VQ-VAE) model on sequence data to quantify the similarity between a given sequence and human VH or camelid VHH domains. This measure of nanobody or antibody nativeness incorporates sequence features that can be used to guide engineering design^[4].

四、Multi-Parameter Optimization of Humanization:

Computational methods have so far considered humanization in isolation. However, when designing antibody therapeutics, all of the properties to be optimized must be considered comprehensively. Immunogenicity arises not only from the intrinsic properties of the antibody itself, but may also derive from developability-related factors, including solubility, aggregation propensity, cross-reactivity, and product heterogeneity caused by insufficient chemical or thermal stability. Multi-parameter optimization problems may also involve patient- or treatment-related factors that influence therapeutic outcomes. For example, a patient's immune status — such as chronic infection or B cell depletion — may affect the likelihood of anti-drug antibody (ADA) formation. The use of computational techniques to simultaneously optimize or screen for antibodies (or nanobodies) with favorable drug-like properties is a powerful strategy for reducing the attrition rate of therapeutic candidates and accelerating the antibody design process. Camelid CDR3 loops were grafted onto a library of humanized VHH frameworks, and camelid CDR1 and CDR2 loops were introduced in various combinations. The sequences in the library were then screened for developability attributes, and their biophysical properties were tested experimentally. In follow-up work, the library-based approach was combined with an LSTM model trained on promising sequences selected from the library, thereby designing new nanobodies that are both human-like and developable.

五、Summary:

Humanization is the principal technique for reducing the immunogenicity of therapeutics derived from non-human sources. At present, this goal is achieved mainly through conventional experimental strategies such as chimerization and CDR loop grafting, combined with the necessary back mutations. However, with the growing wealth of sequence and structural data, it is now possible to develop machine-learning-based humanization scoring and humanization tools. These tools can provide case-specific engineering strategies for each input variable region sequence and are beginning to deliver encouraging results.

TekBiotech (Tianjin) Co., Ltd. has established a nanobody discovery platform based on phage display and yeast display technologies to meet the antibody lead discovery needs of customers worldwide. After preliminary developability assessment of the antibody drug candidate sequences identified by screening, we provide further assured antibody humanization services for the resulting candidate antibody sequences, offering strong technical support for our customers' antibody drug development.

References:

[1] Gordon GL, Raybould MIJ, Wong A and Deane CM (2024) Prospects for the computational humanization of antibodies and nanobodies. Front. Immunol. 15:1399438.

[2] Vincke C, Loris R, Saerens D, Martinez-Rodriguez S, Muyldermans S, Conrath K. General strategy to humanize a camelid single-domain antibody and identi?cation of a universal humanized nanobody scaffold. J Biol Chem. (2009) 284:3273–84.

[3] Sang Z, Xiang Y, Bahar I, Shi Y. Llamanade: An open-source computational pipeline for robust nanobody humanization. Structure. (2022) 30:418–29.e3.

[4] Ramon A, Ali M, Atkinson M, Saturnino A, Didi K, Visentin C, et al. Assessing antibody and nanobody nativeness for hit selection and humanization with AbNatiV. Nat Mach Intelligence. (2024) 6:74–91.

Previous:Stop Sticking with Hybridoma! A Primer on the Three Generations of Antibody Discovery Technology Next:None!