Machine Learning Approaches to Characterization and Root-Cause Analysis of Particulates Formed During Protein Formulation Fill-Finish Operations

In this project, machine learning approaches using convolutional neural networks will be used to extract useful information from collections of images of particles measured after biopharmaceutical fill-finish operations.

Project status

100% Completed

Industry Need

Therapeutic protein drugs are subjected to numerous stresses throughout their manufacturing, storage and shipping that may result in subvisible particles being formed.

Solution

Convolutional neural networks can be used to classify particles generated to mAbs by common manufacturing stresses.

Outputs/Deliverables

Image set development for normal and stressed conditions to create a library to support root-cause analysis.
Applied machine learning approach using convolutional neural networks to extract information from particles measured after biopharmaceutical fill-finish operations.

Impacts

Machine learning approaches using convolutional neural networks will be used to extract useful information from collections of images of particles measured after biopharmaceutical fill-finish operations

Publications

Greenblott, D. N., Wood, C. V., Zhang, J., Viza, N., Chintala, R., Calderon, C. P., & Randolph, T. W. (2024). Supervised and unsupervised machine learning approaches for monitoring subvisible particles within an aluminum-salt adjuvanted vaccine formulation. Biotechnology and Bioengineering, 121(5), 1626-1641. https://doi.org/10.1002/bit.28671

Additional Project Information (Members Only)

Updates
Value Proposition
Related Publications
Deliverables

Not yet a member? Learn more about which level of NIIMBL membership is right for you and your organization.

Project Lead

Regents of the University of Colorado (Boulder)

Participating Organizations

Merck Sharp & Dohme LLC

SentrySciences LLC