Latent variable analysis of Raman spectra for structural characterization of proteins

Abstract

Latent variable analysis of deep UV resonance Raman spectra was demonstrated to be a powerful tool for characterizing protein secondary structural composition. Non-negative independent component analysis (ICA) and pure variable methods, such as stepwise maximum angle calculation (SMAC) and simple-to-use interactive self-modeling mixture analysis (SIMPLISMA), were employed for examination of 10 deep UV Raman (DUVRR) spectra of lysozyme obtained at various stages of its partial denaturation, the first stage of amyloid fibril formation. The non-negative ICA allowed for extracting the spectrum of the β-sheet from deep UV resonance Raman spectra of lysozyme while principle component analysis (PCA) and multivariate curve resolution (MCR) could not separate the β-sheet constituent as an individual component. No initial guess about the features of the β-sheet spectrum was used. Pure variable methods SMAC and SIMPLISMA were found to resolve three independent spectral components assigned to β-sheet, random coil, and native lysozyme.