Therapeutic proteins are exposed to stress during the manufacturing process and can aggregate to form particles. These protein aggregate particles can induce adverse immune responses in patients. This paper discusses the use of image compression techniques proven in facial recognition software to identify and characterize aggregate particles in protein drugs. Using flow-imaging microscopy images, the author creates two-dimensional feature vectors or “fingerprints” of the drug sample. The author also discusses potential applications for this technology. For instance, fingerprints could test how formulation decisions (pH, excipient concentration, etc.) affect protein aggregate populations. Or increase the sensitivity of the algorithm to identify subtle particle morphology differences in drug formulations. Or decrease the sensitivity of the algorithm to detect batch-to-batch production variations.
The full technical paper is embedded below.Daniels-Calderon-Randolph-July-2020