Statistical Methodology for Characterization of Macromolecular Similarity

As with any drug, approval of generic versions of macromolecular drugs requires rigorous evaluation of therapeutic equivalence to the reference drug in order to assure similar efficacy and safety. Compared with drugs that are small organic molecules, the chemical composition and characteristics of macromolecular drugs—typically proteins or polysaccharides—are inherently more variable because of the way these molecules are produced and isolated. It would be advantageous to have a way to determine molecular similarity and, by implication, equivalence without using costly in vivo testing in animals and humans.
The specific aim of the proposal is to develop and test a robust data-driven statistical methodology for assessing similarity among distinct samples of therapeutic macromolecules, whether from different batches or altered processes, or even if produced by different entities entirely. The methodology is based on a genetic algorithm designed to extract relevant features from large, complex datasets, including an assortment of high-resolution mass-spectrometry methods, high-field nuclear magnetic resonance spectroscopy analyses, and several other spectroscopic and chromatographic methods used to characterize macromolecules in solution.