Particle Analysis of Biotherapeutics in Human Serum Using Machine Learning

In recent years, an increasing number of studies assessed the stability of biotherapeutics in biological fluids. Such studies aim to simulate the conditions encountered in the human body and investigate the in vivo stability under in vitro conditions. However, on account of complexity of biological fluids, standard pharmaceutical methods are poorly suited to assess the stability of biotherapeutics. In this study, a fluorescent-labeled therapeutic immunoglobulin G (IgG) was analyzed for proteinaceous particles after mixing with human serum and after incubation at 37°C for 5 days. Samples were analyzed using standard pharmaceutical methods (light obscuration and dynamic imaging). Moreover, we developed a fluorescence microscopy method allowing to semiquantitatively detect IgG particles in serum. Several hundred IgG particles were detected after exposure to serum. Moreover, particle counts and particle size increased in serum over time. The results showed that an IgG may form particles on mixing with serum and novel methods such as fluorescence microscopy are required to gain insight on the stability of biotherapeutics in biological fluids. Furthermore, we showed distinct advantages of machine learning over traditional threshold-based methods by analyzing microscopy images. Machine learning allowed simplifying particles in regards to count, size, and shape.