Molecular Biosimilarity—An AI-Driven Paradigm Shift

Abstract

Scientific, technical, and bioinformatics advances have made it possible to establish analytics-based molecular biosimilarity for the approval of biosimilars. If the molecular structure is identical and other product- and process-related attributes are comparable within the testing limits, then a biosimilar candidate will have the same safety and efficacy as its reference product. Classical testing in animals and patients is much less sensitive in terms of identifying clinically meaningful differences, as is reported in the literature. The recent artificial intelligence (AI)-based protein structure prediction model, AlphaFold-2, has confirmed that the primary structure of proteins always determines their 3D structure; thus, we can deduce that a biosimilar with an identical primary structure will have the same efficacy and safety. Further confirmation of the thesis has been established using technologies that are now much more sensitive. For example, mass spectrometry (MS) is thousands of times more sensitive and accurate when compared to any form of biological testing. While regulatory agencies have begun waiving animal testing and, in some cases, clinical efficacy testing, the removal of clinical pharmacology profiling brings with it a dramatic paradigm shift, reducing development costs without compromising safety or efficacy. A list of 160+ products that are ready to enter as biosimilars has been shared. Major actions from regulatory agencies and developers are required to facilitate this paradigm shift.