Selected Grantee Publications

Systematic Multi-trait AAV Capsid Engineering for Efficient Gene Delivery

Eid et al., Nature Communications. 2024.

https://doi.org/10.1038/s41467-024-50555-y

Engineering novel functions into proteins while retaining desired traits is a key challenge for developers of viral vectors, antibodies, and inhibitors of medical and industrial value. In this study, investigators developed Fit4Function, a generalizable machine learning (ML) approach for systematically engineering multi-trait adeno-associated virus (AAV) capsids. Fit4Function was used to generate reproducible screening data from a capsid library that samples the entire manufacturable sequence space. The Fit4Function data were used to train accurate sequence-to-function models, which were combined to develop a library of capsid candidates. Compared to AAV9, top candidates from the Fit4Function capsid library exhibited comparable production yields; more efficient murine liver transduction; up to 1,000-fold greater human hepatocyte transduction; and increased enrichment in a screen for liver transduction in macaques. The Fit4Function strategy enables prediction of peptide-modified AAV capsid traits across species and is a critical step toward assembling an ML atlas that predicts AAV capsid performance across dozens of traits. Supported by ORIP (P51OD011107, U42OD027094), NIDDK, NIMH, and NINDS.