Quantifying protein residues in APIs: Bradford assay mechanism and limitations, outperformed by HILIC-MS/MS

The molecular complexity of active pharmaceutical ingredients (APIs) has increased over recent decades, with production largely relying on microbial processes such as fermentation. However, these processes introduce non-therapeutic protein impurities. Accurate quantification of protein impurities in APIs is critical, as underestimation risks safety and efficacy, while overestimation leads to costly over-purification. Here, we evaluated five conventional protein assays with diverse standards and found their accuracy compromised by protein structural diversity. To deconstruct the mechanism of this limitation, we examined the Bradford assay as a case study. Machine learning models trained on data from over 140 standards revealed that, beyond the known roles of basic and aromatic residues, other amino acids exert positive or negative effects on the Bradford assay response-as predicted by Lasso regression and validated with five designed peptides. Together with molecular docking and interaction analysis, which identified electrostatic interactions as the primary driving force, these findings collectively account for the inherent bias of such structure-dependent assays. The systematic evaluation and mechanistic insight prompted us to develop a hydrophilic interaction liquid chromatography coupled with tandem mass spectrometry (HILIC-MS/MS) method for amino acid analysis, enabling accurate protein quantification at ppm levels in matrices. Applied to eight commercial APIs, the method revealed substantial variation in protein residue levels (9.53-5570 ppm). By providing precise measurement of protein impurities in small-molecule APIs, this study helps ensure drug safety and efficacy at no excessive purification costs, while also offering insights into the molecular mechanisms underlying the Bradford assay.