Comparing Analytical Methods via Predictive ModellingPerme, Tinkara (Avtor) Hajdinjak, Melita (Mentor) Porenta, Tine (Komentor) exploratory factor analysis (EFA)principal component analysis (PCA)multiple linear regressionanion exchange chromatography (AEX)multi-attribute method (MAM)Complex biological drugs require thorough characterization throughout development and manufacturing to define their structure, detect structural changes, and identify process deviations. Multiple analytical techniques are used for this purpose. Recently, a newer method called multi-attribute method (MAM) has emerged that enables more holistic protein characterization and could replace older, already established methods, one of them being anion exchange chromatography (AEX). To evaluate whether AEX can be replaced by MAM, methods must be compared to determine if they provide equivalent information. We therefore developed a predictive model that estimates the sum of acidic peaks measured by AEX using more comprehensive measurements obtained by MAM, including glycosylation patterns, amino acid modifications, protein degradation levels, and glycation. Because MAM yields a complex dataset (over 50 variables with substantial collinearity), dimensionality reduction was necessary prior to modelling. We did so in an interpretable way to ensure that the magnitude and direction of coefficients could also be interpreted from the chemical perspective. We applied and compared two different dimensionality reduction techniques: exploratory factor analysis (EFA) and principal component analysis (PCA), which are widely used but often poorly understood. Both approaches produced similar results. In the glycosylation subset, approximately 80% of the variance was explained by the first two principal components from PCA, reflecting the level of X component content on the glycans and glycan size, which could not be predicted from the degree of X components alone. Incorporating these variables into a multiple regression model, alongside other MAM-derived protein modifications, showed that oxidations, X components on glycans, glycan size, specific protein clippings (degradations), and glycation increase the sum of acidic peaks measured by AEX. The provided characteristics explained approximately 73% of the variance in the sum of acidic peaks.20262026-01-08 15:00:02Magistrsko delo/naloga177823VisID: 63091COBISS_ID: 265400579sl