Optimal trade-off control in machine learning–based library design - with application to adeno-associated virus (AAV) for gene therapy

Aug 6, 2026

Upload by Anda

Adeno-associated viruses (AAVs) are widely used as delivery vectors for gene therapies, but many naturally generated capsid libraries include a large fraction of nonfunctional variants and often lack targeted, efficient delivery. This work presents a machine learning method for designing AAV peptide insertion libraries with substantially improved packaging fitness while maintaining diversity. Results show about fivefold higher packaging fitness than the standard NNK library and roughly 10-fold more successful variants after selection for human brain infection, enabling promising glial-specific candidates and a design framework extendable beyond AAV.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages17

Reading Time43 min

AccessFree to download

Abstract

Introduction

Directed evolution and AAV libraries

Computational design of library parameters

UC Berkeley UC Berkeley Previously Published Works Title Optimal trade-off control in machine learning-based library design, with application toadeno-associated virus (AAV) for gene therapy. Permalink [https://escholarship.org/uc/item/87b6k6z8](https://escholarship.org/uc/item/87b6k6z8) Journal Science Advances, 10(4) Authors Zhu, Danqing Brookes, David Busia, Akosuaet al. Publication Date 2024-01-26 DOI 10.1126/sciadv.adj3786 Peer reviewed [eScholarship.org](eScholarship.org) Powered by the California Digital Library University of California MOLECULAR BIOLOGY Optimal trade-off control in machine learning–based library design, with application to adeno-associated virus (AAV) for gene therapy Danqing Zhu1†‡, David H. Brookes2†, Akosua Busia3†, Ana Carneiro4, Clara Fannjiang, Galina Popova5,6,7, David Shin5,6,7, Kevin C. Donohue6,8,9,10, Li F. Lin4, Zachary M. Miller11, Evan R. Williams11, Edward F. Chang12, Tomasz J. Nowakowski5,6,7,10,12, Jennifer Listgarten3,13*, David V. Schaffer1,4,14,15,16,17* Adeno-associated viruses (AAVs) hold tremendous promise as delivery vectors for gene therapies. AAVs have been successfully engineered—for instance, for more efficient and/or cell-specific delivery to numerous tissues—by creating large, diverse starting libraries and selecting for desired properties. However, these starting libraries often contain a high proportion of variants unable to assemble or package their genomes, a prerequisite for any gene delivery goal. Here, we present and showcase a machine learning (ML) method for designing AAV peptide insertion libraries that achieve fivefold higher packaging fitness than the standard NNK library with negligible reduction in diversity. To demonstrate our ML-designed library’s utility for downstream engineering goals, we show that it yields approximately 10-fold more successful variants than the NNK library after selection for infection of human brain tissue, leading to a promising glial-specific variant. Moreover, our design approach can be applied to other types of libraries for AAV and beyond. copyright © 2024 the Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. no claim to original U.S. Government Works. distributed under a creative commons Attribution noncommercial license 4.0 (cc BY-nc) . INTRODUCTION Adeno-associated viruses (AAVs) hold major promise as delivery vectors for gene therapy. While naturally occurring AAVs can be clinically administered safely and in some cases efficaciously, they have a number of shortcomings that limit their use in many human therapeutic applications. For example, naturally occurring AAVs do not target delivery to specific organs or cells, their delivery efficiency is limited, and they are susceptible to preexisting neutralizing antibodies (1–3) . Consequently, directed evolution of the AAV capsid protein has emerged as a powerful strategy for engineering therapeutically suitable or optimal AAV variants. In directed evolution, a diversified library of AAV capsid sequences is subjected to multiple 1california institute for Quantitative Biosciences, University of california, Berkeley, Berkeley, cA 94720, USA. 2Biophysics Graduate Group, University of california, Berkeley, Berkeley, cA 94720, USA. 3department of electrical engineering and computer Sciences, University of california, Berkeley, Berkeley, cA 94720, USA. 4department of chemical and Biomolecular engineering, University of california, Berkeley, Berkeley, cA 94720, USA. 5department of Anatomy, University of california San Francisco, San Francisco, cA 94143, USA. 6department of Psychiatry and Behavioural Sciences, University of california San Francisco, San Francisco, cA 94143, USA. 7eli and edythe Broad center for Regeneration Medicine and Stem cell Research, University of california San Francisco, San Francisco, cA 94143, USA. 8School of Medicine, University of california San Francisco, S

FAQs

What problem does the document address in AAV library design?

Starting AAV libraries often contain many variants that cannot assemble or package their genomes, and naturally occurring AAVs also face limitations in targeting and delivery efficiency. This reduces effectiveness for gene therapy goals.

How does the proposed method improve over the standard NNK library?

The machine learning approach designs AAV peptide insertion libraries with about fivefold higher packaging fitness than the standard NNK library while causing negligible reductions in diversity.

What downstream benefit is demonstrated for the ML-designed library?

After selection for infection of human brain tissue, the ML-designed library yields approximately 10-fold more successful variants than the NNK library, including a promising glial-specific variant.

Aug 6, 2026

Upload by Anda