CDRH3 flanking residues more conserved and easier to predict than those in the center

Summary

Residues at the ends of CDRH3 are more conserved than those near the center of the loop (1). As a result, they are easier to predict, as shown by Paul et al. (2) and (3).

Figures

Figure 1 from (2)

See also

• Optimal scaling of antibody LMs improves prediction of masked CDRH3 residues but not framework residues

1.

Guest JD, Vreven T, Zhou J, Moal I, Jeliazkov JR, Gray JJ, et al. An expanded benchmark for antibody-antigen docking and affinity prediction reveals insights into antibody recognition determinants. Structure. 2021;29(6):606-621.e5. Available from: https://doi.org/10.1016/j.str.2021.01.005

2.

Paul S, Huang Y, Marks D. An energy based model for incorporating sequence priors for target-specific antibody design. In: NeurIPS 2023 Generative AI and Biology (GenBio) Workshop. 2023. Available from: https://openreview.net/forum?id=Ejw5zOOgLp

3.

Neyestanak MS, Burbach SM, Ng K, Gangavarapu P, Hurtado J, Magura J, et al. Data-optimal scaling of paired antibody language models. openRxiv; 2025. Available from: https://doi.org/10.1101/2025.09.02.673765