Case Study
AI/ML Optimization of Coding Sequences Reduces Genome Truncations in AAV Vectors
The coding sequence (CDS) of a gene-of-interest (GOI) in AAV gene therapies provide little opportunity for optimization besides evolutionarily optimized codon frequency usage. However, if the CDS of a GOI has a risk for causing AAV genome truncations, it can lead to significant manufacturing challenges and provide limited options for subsequent bioprocessing improvement. Here, we explored using AI/ML to optimize the CDS with the aim of reducing truncation events in this genomic element.
We used FORMsightAI to predict the truncation risk of two vector designs, AAV5_P1_PRO1_GOI containing a reporter gene with a P1 promoter and P1_PRO1_GOI containing a housekeeping gene with a P1 promoter (Table 1). We then optimized the CDS regions using FORMsightAI to create AAV5_P1_PRO1_GOI_OPT and P1_PRO1_GOI_OPT and then evaluated them using long-read sequencing.
Although the coding sequences of AAV5_P1_PRO1_GOI (Figure 1) and P1_PRO1_GOI (Figure 2) had few truncation issues with limited opportunity for improvement (purple dots), FORMsightAI optimization still achieved a notable reduction in truncation events for both GOIs in AAV5_P1_PRO1_GOI_OPT and P1_PRO1_GOI_OPT (pink dots). This resulted in truncation reductions of 14% and 30% for AAV5_P1_PRO1_GOI and P1_PRO1_GOI, respectively, which can have a significant impact in manufacturing.
