Twist Bioscience HQ
681 Gateway Blvd
South San Francisco, CA 94080
Many tools for codon optimization follow a rules-based approach, such that individual codons are optimized in isolation. While useful, this approach ignores the higher-order effects that are produced by codon combinations. It is therefore likely that rules-based codon optimization fails to meet the full potential output for any given gene. Here, we set out to test whether Twist’s LLM-driven codon optimization approach, which is able account for combinatorial codon effects, is capable of outperforming legacy rules-based and machine learning approaches across a panel of 32 antibody constructs.
Covered in this White Paper:
✓ The benefits of LLM-based codon optimization algorithms for increasing protein expression
✓ 6 codon optimization algorithms tested for their impact on protein expression across a panel of 32 antibodies
✓ Analysis of underlying factors driving higher performance observed with LLM-based algorithms