Structural Similarity of Biological Drugs Using Statistical Signal Processing and Nuclear Magnetic Resonance Spectral Pattern Analysis.

Biosimilar drugs are highly similar to the available marketed drugs and have no clinically meaningful differences in terms of safety, purity, and potency. As per stringent drug regulatory requirements, biosimilar drugs must match closely to all attributes of the listed marketed drug, including establishing high similarity of higher-order structures. Here, we have developed a combined approach using high-resolution two-dimensional nuclear magnetic resonance (NMR) spectra and image-based statistical signal processing algorithms to establish robust comparability of critical quality attributes of biological drugs. We have integrated a computational approach to 2D NMR data analysis, which could replace the traditional methods of manually extracting chemical shift values and intensities for each peak and performing a range of statistical analyses, which are laborious and prone to ambiguity. Our algorithm simplifies and streamlines this process, making it more accurate, less time-consuming, and avoiding personal biases. We have employed our methods with a diverse range of biotherapeutics and complex NMR data and shown a degree of similarity between reference and test drugs with our differentially assigned similarity scores.