(M1530-04-21) From Sequence to Binding Clones, a Novel Computational Platform for Nanobody Discovery

Purpose: Nanobodies are small single-domain camelid antibodies with high specificity and stability, making them promising therapeutic antibody modality. However, their discovery is often time-consuming and costly. Leveraging high-throughput Next Generation Sequencing (NGS), researchers nowadays might acquire millions of nanobody sequences for in silico therapeutic nanobody screening.

Methods: Here, we developed a computational platform for the in-depth analysis of nanobody repertoires and efficient selection and prioritization of antigen-specific nanodiodes from the post-immune NGS sequence data. As the initial proof of concept, we applied this platform to computationally select anti-protein X nanobodies from an immunized llama B cell library.

Results: Both predicted binding and non-binding VHH clones were deliberately selected from the output of the platform and subjected to vigorous in vitro validation. The validation process confirmed selected VHH clones’ classification as either affirmative binders or non-binders of target protein X. Additionally, our platform streamlines the iterative selection of candidates for the subsequent round of affinity matured and optimized clones through exclusive digital screening.

Conclusion: Overall, this study presents an innovative, efficient, and cost-effective framework for accelerating nanobody discovery, combining next generation sequencing (NGS) and large protein language model (PLM) to streamline the identification and validation of therapeutic nanobodies.

References: [1] Lyu, X., et al. (2022). "The global landscape of approved antibody therapies." Antibody Therapeutics 5(4): 233-257.
[2] Yaari, G. and S. H. Kleinstein (2015). "Practical guidelines for B-cell receptor repertoire sequencing analysis." Genome Medicine 7(1): 121.
[3] Lin, Z., et al. (2023). "Evolutionary-scale prediction of atomic-level protein structure with a language model." Science 379(6637): 1123-1130.
[4] Madani, A., et al. (2023). "Large language models generate functional protein sequences across diverse families." Nat Biotechnol 41(8): 1099-1106.
[5] Zhang, W., et al. (2023). "Developability assessment at early-stage discovery to enable development of antibody-derived therapeutics." Antib Ther 6(1): 13-29.

Figure 1. Animal immunization and B cells enrichment. (A) A naïve alpaca was immunized with human Protein X S.C. according to the schedule above (Figure 1). Bleed samples were collected at the indicated time point. At the end of study, alpaca PBMCs were prepared from 300 ml whole blood. (B) Total Alpaca IgG and VHH specific titer of Protein X were evaluated with appropriate detection reagent through out the immunization. (C) Collected immunized PBMCs were stained with biotin-protein X, eventually ~30K VHH specific B cells were bulk sorted into collection tubes.



Figure 3. In vitro validation of in silico selected VHH clones 20 different VHH clones were selected based on clone frequency (dark green), evolutionary phylogenetics analysis (light green) and the combination (white). (Left) The binding affinity of selected VHH against protein X were evaluated in HT in-direct ELISA assay against protein X. (Right) The EC 50 values of all tested clones were ranked from lowest to highest


Figure 4. Affinity matured VHH clones by PLM model. PLM model was utilized for the in silico affinity maturation, and 20 mutants were recommended for subsequent in vitro HT ELISA. The best performing clones from platform selection (top panel) and PLM model prediction (bottom panel) were further evaluated in SPR analysis. The K disassociation were significantly improved in the later session.