CDR Diversification Now Available: One Great Antibody, Unlimited Design Possibilities

Traditional CDR diversification relies largely on random mutagenesis or semi-rational design. While simple to implement and capable of generating many sequences quickly, its limitations have grown more apparent as R&D requirements rise.

1. Inefficient sequence space exploration
Random mutagenesis produces large numbers of non-functional variants, burying truly valuable sequences within a vast pool of candidates.

2. Structural stability is hard to guarantee
Mutations without structural constraints can disrupt antibody folding, causing expression difficulties or aggregation.

3. The trade-off between library size and quality
Achieving acceptable hit rates requires screening millions or even billions of variants, making cost and cycle time a bottleneck.

Ultimately, what R&D truly needs is a smaller, higher-quality, structurally sound candidate library.

Traditional CDR diversification resembles random exploration of sequence space. Researchers typically specify which CDRs are involved in mutation, then rely on random combinations to generate large numbers of sequences, followed by gradual experimental filtering.

Structure-guided CDR diversification, by contrast, is a design process better suited to human-AI collaboration:

R&D experts: turning experience into design constraints
Experts contribute their project experience to the design strategy — specifying which positions must be preserved, which can vary, and the range of parental similarity to maintain.

AI: structure-based directed sampling
AI integrates the antibody's three-dimensional structure to systematically analyze each CDR position, automatically identifying key residues involved in antigen binding, structural support sites, and regions more suitable for introducing diversity. It further evaluates the structural perturbations from different mutant combinations, performs directed sampling of sequence space within design constraints, and generates a set of candidate sequences that balance structural soundness with sequence diversity.

Through human-AI collaboration, random exploration is transformed into bounded, evidence-based design — allowing more experimental resources to focus on candidates most worth validating.

Recognizing that different projects have different data foundations, we provide two design modes to meet the needs of different R&D stages.

1. Sequence Mode : Suitable for projects that only have antibody and antigen sequences. The system automatically performs CDR annotation, structure prediction, and subsequent diversity design — no manual modeling required. This mode enables rapid library design at the early project stage. 2. Structure Mode : Suitable for projects that already have an antibody–antigen complex structure (crystal structure or high-quality predicted model). The system performs analysis directly from the existing structure, enabling more refined position-level structural evaluation and CDR diversity design.

The system combines structural, evolutionary, and physicochemical information for sequence design, and controls critical residues through position-level protection strategies — ensuring diversity while maintaining structural soundness, and yielding higher-quality candidate sequences ready for library construction.

Compatible with major numbering schemes including Kabat, Chothia, IMGT, and AHO, enabling consistency with different databases, literature, and workflows without extra conversion.

The entire computational workflow completes in approximately 2 hours, outputting a candidate library ready for direct experimental screening.

To validate the real-world effectiveness of CDR diversity design, we used a Trop2 antibody as the parental molecule, constructed a diversity library under the condition that all CDR sequence identities were below 70% relative to the parent, and tested representative candidates for expression, purification, and biophysical properties.

1. Sequence identity: CDR1/2/3 sequences of both heavy and light chains showed less than 70% identity to the parental antibody. 2. Affinity: 2 antibodies maintained more than 50% of parental binding affinity. 3. Expression: 11 antibodies achieved expression levels above 50% of the parental antibody. 4. Purity: 8 antibodies showed better SEC-HPLC purity than the parental antibody.

The experimental results confirm that CDR diversification does not have to mean compromised performance. Through structure-guided design, it is possible to obtain candidate antibodies with both high diversity and strong developability potential, even under low sequence identity conditions.

This is a question many R&D scientists ask. While both involve CDR regions, the goals are entirely different. They address different stages and objectives — but can also be used together in the same project.

1. CDR Diversity Engineering
Focuses on systematically designing a broader set of sequence combinations around the same parental antibody, expanding sequence space to provide more options for downstream experimental screening.

2. Affinity Maturation
Focuses on targeted optimization of an existing antibody — improving binding to the target antigen through a limited number of key mutations.

CDR diversity engineering is designed to build a richer set of candidate sequences, and is especially well-suited for the following scenarios:

1. You have obtained a lead antibody and want to further expand the sequence space.
2. You want to construct a diverse CDR library around the same parental antibody.
3. You want more candidate molecules with the same recognition features but different sequences.
4. You want to provide more sequence options for downstream experimental screening.
5. You want to build a sequence family with differentiated characteristics.