Case Study

AI/ML is Able to Reliably and Reproducibly predict AAV Genome Truncation

DNA icon

Context

AAV genome truncations are a major challenge in gene therapy development, reducing production lot yields and increasing manufacturing costs. The ability to reliably and reproducibly predict the propensity of vector designs for producing truncated genomes early in discovery can mitigate manufacturing challenges later in development, saving significant time and cost. Here, we investigated whether AI-based predictions of AAV genome truncation propensity reliably and reproducibly correlate with long-read sequencing data.

methods icon

Methods

We used FORMsightAI to predict the truncation propensity of 42 AAV vectors and then used PacBio long-read sequencing to sequence the genomes of the vectors. We then compared the FORMsightAI-predictions with long-read sequencing data to evaluate the reliability and reproducibility of the predictions.

splitting molecule icon

Results

We observed a strong correlation between FORMsightAI predictions for genome truncation propensity and long-read sequencing of the vectors (Figure 1). Furthermore, an overlay of the FORMsightAI predictions with long-read sequencing data for the individual 42 vectors demonstrated that the predictions were consistent and accurate with the actual sequencing data (Figure 2).

Figure 1 image
Figure 1: Correlation plot of predictions with current truncation model vs NGS data. Dashed line is the 1:1 slope of perfect correlation. All data are plotted on a single graph in a log scale. Blue dots represent the data points used for training the model, while red dots indicate the held-out data points.
Figure 2 image
Figure 2: We observed a strong correlation between FORMsightAI predictions for genome truncation propensity and long-read sequencing of the vectors (Figure 1). Furthermore, an overlay of the FORMsightAI predictions with long-read sequencing data for the individual 42 vectors demonstrated that the predictions were consistent and accurate with the actual sequencing data (Figure 2).

Impact

FORMsightAI is a powerful prediction tool that can reliably and reproducibly predict AAV genome truncation.

You may also be interested in...

AI/ML Optimization Reduces AAV Discovery Timeline from Years to Days

Read more

Optimization of the CDS Using an AI/ML Model Outperforms an Open-Source Codon Optimization Tool by Improving Enzymatic Activity by 20%

Read more

AI/ML Optimization of Coding Sequences Reduces Genome Truncations in AAV Vectors

Read more

Get Your Therapeutic Candidate to Market Faster

TALK TO OUR EXPERTS
Schedule a Discovery Call