Case Study
Reducing the Risk of Immunogenicity Using an AI/ML Model
An immunogenic response to a drug can lead to a severe adverse event in a patient. Therefore, it is crucial to remove any immunogenic risk factors during the pre-clinical development stage to ensure patient safety. CpG islands are an important risk factor for AAV vectors, because they can trigger an immune response through TLR9 signaling. Here, we evaluated whether CDS optimization through our AI/ML model reduces the %CpG island peaks across the CDS of a gene.
We employed FORMsightAI simulation to evaluate the immunogenicity risk of P1_PRO1_GOI containing a housekeeping gene with a P1 promoter (Table 1) by looking at %GC content and %CpG islands. We then optimized the CDS using FORMsightAI to create P1_PRO1_GOI_OPT. The constructs were manufactured in partnership with a major CDMO, and the vector designs were evaluated for genomic identity using long-read sequencing.
Generally, codon optimization with open-source tools can lead to an increase in CpG dinucleotides, resulting in an increase in GC content. As a part of our optimization process, we aimed to maintain the GC content consistent. Optimization of the CDS of a housekeeping gene using FORMsightAI did not lead to a significant change in %GC content (Figure 1; top panel). Additionally, we were able to reduce the %CpG peaks in the original CDS (purple) following FORMsightAI optimization (pink) (Figure 1; bottom panel), thereby reducing the risk of immunogenicity mediated by TLR9.
