AI
AI

Predictive Machine Learning Model for Evaluating AAV Capsid Fitness in Gene Therapy Applications

Photo credit: phys.org

New Machine Learning Model Enhances AAV Capsid Fitness for Gene Therapy

A recent study published in Human Gene Therapy introduces an innovative machine learning model designed to streamline the process of developing adeno-associated virus (AAV) capsids. This model promises to serve as an effective alternative to extensive in vitro experimentation, aimed at improving the fitness of AAV capsids to facilitate more cost-effective gene therapies.

Improving the yield and overall fitness of AAV capsids is crucial for lowering production costs, thereby making gene therapies more accessible to patients. The research, conducted by Christian Mueller and his team at Sanofi, highlights a contemporary machine learning framework that predicts the fitness of AAV2 capsid mutants based solely on their amino acid sequences.

“Our model, which integrates a protein language model with traditional machine learning methodologies, demonstrated an impressive prediction accuracy with a Pearson correlation coefficient of 0.818 concerning capsid fitness,” the researchers noted. “Moreover, when our model was tested on entirely separate datasets, it displayed considerable robustness and adaptability, even when applied to AAV capsids containing multiple mutations.”

Dr. Thomas Gallagher, Managing Editor of Human Gene Therapy and affiliated with the University of Massachusetts Chan Medical School, commented on the significance of this advancement: “The rise of AI-driven techniques represents a promising shift in capsid engineering. Such methods could offer a more systematic, thorough, and cost-effective alternative compared to traditional directed evolution and rational design approaches. The study led by Wu et al. marks an important progression in developing artificial intelligence tools within the gene therapy domain.”

More information: Jason Wu et al, Prediction of Adeno-Associated Virus Fitness with a Protein Language-Based Machine Learning Model, Human Gene Therapy (2025). DOI: 10.1089/hum.2024.227

Citation: Machine learning model to predict the fitness of AAV capsids for gene therapy (2025, April 21) retrieved from https://phys.org/news/2025-04-machine-aav-capsids-gene-therapy.html

This article is for informational purposes only and is subject to copyright. Any reproduction is prohibited without permission.

Source
phys.org

Related by category

Sexism Disrupts Emotional Synchrony and Undermines Team Performance

Photo credit: phys.org Gender Bias in the Workplace: Its Impact...

Using Humor in Communication to Foster Connections and Build Trust Among Scientists

Photo credit: phys.org The Impact of Humor in Scientific Communication Researchers...

Physicists Examine Quantum Theory Using Atomic Nuclei from Nuclear Reactions

Photo credit: phys.org Advancements in Understanding Atomic Nuclei Magnetism Many atomic...

Latest news

11 Stylish Outdoor Bar Carts to Elevate Your Summer Experience in 2025

Photo credit: www.architecturaldigest.com Dimensions: 22"W x 23"D x 36"H Offering a...

Roundtrip Flights from Phoenix to Amsterdam, Netherlands: $554 (Basic Economy) / $754 (Regular Economy) – All Taxes Included!

Photo credit: www.theflightdeal.com Traveling to Amsterdam has become more affordable...

Voicing Our Values: AAC&U Annual Conference Inspires Attendees to ‘Reclaim Liberal Education’

Photo credit: www.higheredjobs.com Contimis Works/ Shutterstock During a recent gathering of...

Breaking news