Develop a model that can better predict functional titer from lower-cost methods of analysis compared to using transmission electron microscopy (TEM) images of viruses
Carnegie Mellon University and Sartorius aim to develop a computational workflow to integrate results of rapid viral titer assays to best match the outcomes of high precision determination of functional titer.
The two organizations will incorporate and validate their methods by using machine learning (ML) approaches.
Experiments will include:
This project will determine the workflow that optimizes analytical assay reproducibility and repeatability and will predict gaps in knowledge that may limit product analysis.
Create computational assessment of rapid viral titer assays (e.g., qPCR, ddPCR, optical imaging, probe-capture methods) to best predict functional titer of viruses
Provide evaluation methods that are applicable during viral production (e.g., bioreactor stage) to facilitate early decision making
Decrease decision-making time and provides reduced use of costly assays
Implementing machine-guided analytics and qPCR-based workflows for AAV8 capsid quality assessment can significantly reduce manufacturing costs by minimizing batch failures and optimizing production decisions. By achieving 98% accuracy in capsid classification and enabling early prediction of viral titers, manufacturers can save on raw material expenses and improve scalability, translating into faster release timelines and an estimated cost reduction of up to 20–30% per batch through decreased reliance on labor-intensive TEM analysis and improved process efficiency.
Benchmarking of Analytical Methods: Established reproducible qPCR and ELISA workflows for full/empty AAV8 capsid analysis, validated against TEM imaging as the ground truth.
Machine Learning Model Development: Created a convolutional neural network (CNN) capable of predicting capsid composition with up to 98% accuracy, reducing manual image analysis time.
Novel Assay Implementation: Tested live-cell infectivity assay using Sartorius Incucyte and developed micelle-tagged electrophoresis probes for AAV analysis.
Gutierrez, L., & Robinson, A., (2025) An Automated Workflow Leveraging Machine Learning for Physical Titer Determination from Cryo-TEM Images of Adeno-Associated Virus Capsids. Chemical and Biomedical Imaging, https://doi.org/10.1021/cbmi.5c00070
Robinson, A., Conference Participant, 5.2T-129: Machine-guided rapid decision-making on the quality of viruses as products, NIIMBL National Meeting, Washington, D.C., June 26, 2025.
Yang, Y. & Robinson, A. S., Functional and Physical Titer Determination of AAV8 Production, Biotherapeutics and Vaccines Development Gordon Research Conference, Galveston, TX, March 17, 2024.
Login to the NIIMBL member portal to access additional project information, including presentations, progress updates, reports, and more.
Not yet a member? Learn more about which level of NIIMBL membership is right for you and your organization.
Carnegie Mellon University
Sartorius Stedim
