Elevate Antibody Development with
AI-Guided Library Design

Designed using deep learning, the Twist AI Hypermutated scFv Library provides a versatile platform for antibody discovery. The neural network used to design the library mined millions of antibody sequences to generate diverse antibody repertoires in a process that resembles natural B-cell receptor hypermutation and recombination.

The library limits diversity in complementarity-determining regions 1 (CDR1) and 2 (CDR2) while maximizing that of CDR3. Four combinations of heavy chains and light chains (VH3-23/VK1-39, VH3-23/VK3-20, VH1-69/ VK1-39, and VH1-69/VK3-20) each incorporate 200 linked HCDR1-HCDR2 sequences with 100,000 HCDR3s, and 100 linked LCDR1-LCDR2 sequences with 10,000 LCDR3s. Combinatorial assembly results in a fully human scFv library with 400,000 HCDR3s, 40,000 LCDR3s, and a diversity of 1 x 10⁹.

The sequences in the AI Hypermutated scFv library are computationally derived using a carefully selected subset of a full database of naturally occurring human antibodies. For CDR1, CDR2, and CDR3 of IGHV3-23, the length distributions of the selected antibody sequences (purple) closely mimic the natural human antibody repertoire (black).

Go from panning to functional assays in 10–12 weeks. The process starts with phage screening the diverse Twist AI Hypermutated scFv Library against target antigens and ends with reformatting candidate antibody fragments to full-length IgG.

You can also license the AI Hypermutated scFv library to initiate your own in-house discovery projects. To learn more, get in touch at biopharma@twistbioscience.com.

The AI Hypermutated scFv Library effectively uncovers SARS-CoV-2 S1 antibody leads with high binding affinities.

A panel of anti-S1 antibodies from the AI Hypermutated scFv library shows inhibition of S1 binding to ACE2-expressing VERO E6 cells. A flow cytometry plot for representative clone TB268-14 illustrates a shift in the transfectant population compared to the parental population.