Quantitative Comparison and Clustering of Circular Dichroism Spectra Using a Symmetrized Weighted Spectral Difference.

Spectroscopy (UV-visible, circular dichroism, infrared, Raman, fluorescence, etc.) is of fundamental importance to determine the structures of macromolecules and monitor their stability, especially for drug products, based on proteins or nucleic acids. In their 2014 article, Dinh et al. proposed Weighted Spectral Difference (WSD) as a method to quantitatively compute the dissimilarity of a given spectrum to a reference one. Despite the various properties of this method, its lack of symmetry and dependence on the selection of a reference limits the range of possible applications. Here, we propose a reference-free, symmetrized version of WSD (SWSD) that allows the computation of a semi-distance between two spectra. SWSD can be applied to perform group comparisons, track spectral kinetics, or construct a SWSD matrix leading to the hierarchical clustering of spectra. This method was tested on circular dichroism spectra from a split-virus-based (influenza) vaccine and a recombinant spike protein (COVID-19 vaccine). This approach resulted, first, in a perfect clustering of influenza A and B viruses into two distinct clusters, and second, in the detection of the change of secondary structure of the spike protein during a heating experiment, identifying two main temperatures of denaturation (Tm) by SWSD kinetics, in agreement with results obtained by conventional DSC. In summary, we have shown that SWSD is a versatile and efficient tool for quantitative spectral comparison, tracking spectral kinetics and enabling relevant unsupervised classification.