Core Concepts
A diffusion-based generative model can efficiently design diverse and viable rAAV capsid sequences, enabling the exploration of broader mutational landscapes and overcoming limitations of traditional approaches.
Abstract
The study presents an innovative approach to designing and validating recombinant adeno-associated virus (rAAV) capsid sequences using a diffusion-based generative model. The key highlights are:
The diffusion model was trained on publicly available AAV2 data and generated 38,000 diverse AAV2 viral protein (VP) sequences. 8,000 of these were evaluated through biological activity testing, demonstrating a significant improvement in performance compared to traditional methods.
In the absence of AAV9 capsid data, the model was used to directly generate viable AAV9 sequences with up to 9 mutations, outperforming random mutation approaches.
The study explored the mutational landscape of AAV9 hypervariable regions IV, V, and VIII, identifying tolerant regions and unfavorable amino acid substitutions. This provides valuable insights for further capsid engineering.
The generative modeling approach enables the exploration of broader mutational landscapes while maintaining manageable library sizes, addressing the limitations of traditional capsid library design and screening methods.
The research represents a significant advancement in the design and functional validation of rAAV vectors, offering innovative solutions to enhance specificity and transduction efficiency in gene therapy applications.
Stats
"The proportion of viable samples exceeds 90% when the number of mutations ranges from 7 to 20."
"The proportion of viable samples is approximately 80% for mutation numbers ranging from 4 to 6."
"The generated sequences displayed a viability proportion of approximately 50% when the number of mutations ranged from 9 to 10."
Quotes
"This research represents a significant advancement in the design and functional validation of rAAV vectors, offering innovative solutions to enhance specificity and transduction efficiency in gene therapy applications."
"The generative modeling approach enables the exploration of broader mutational landscapes while maintaining manageable library sizes, addressing the limitations of traditional capsid library design and screening methods."