Case Study
Optimization of the CDS Using an AI/ML Model Outperforms an Open-Source Codon Optimization Tool by Improving Enzymatic Activity by 20%
Species-specific codon optimization has been a valuable tool in research over the past decades. While it has been built on empirical data, it has played an important role in optimizing protein expression. However, in the current age of genetic medicines, more advanced tools are needed to address the complexity of designing and manufacturing gene therapies. Here, we compared the performance of our proprietary AI/ML model versus an open-source tool for optimizing the activity of an enzyme.
We evaluated four different vector designs coding for an enzyme as a gene-of-interest (GOI) in this study (Table 1): (1) pAAV9: negative control, (2) pAAV9_FORMsight1_GOI: CDS optimized with method 1, (3) pAAV9_FORMsight2_GOI: AI optimized with method 2, and (4) pAAV9_OPT_GOI: open-source codon-optimized. These constructs were packaged into AAV9 capsids and manufactured in 2L suspension bioreactors in Expi293FTM HEK cells, and evaluated for mRNA expression using ddPCR and enzymatic assay using ELISA by an academic collaborator.
Interestingly, we found that both vector designs whose coding sequences were optimized using FORMsightAI outperformed standard open-source optimization (Table 2). In fact, for both FORMsightAI-optimized versions of the coding sequence, we observed a 20% increase in mRNA transcript levels and enzyme activity with ddPCR and fluorometric enzymatic assay, respectively.
