Selecting Two-Dimensional Cross-Correlation Functions to Enhance Interpretation of Near-Infrared Spectra of Proteins

Abstract

Two-dimensional infrared (2D-IR) cross-correlation can facilitate the interpretation of infrared spectra. However, the fluctuation of noise and the overlapped component bands often complicate the analysis of 2D-IR spectra. The ratio of asynchronous to synchronous correlation functions (F) can serve as a control of coherence for the cross peaks in 2D correlation maps. A synchronous cross peak with an F value close to 0 corresponds to the good coherence of cross-correlated pair of signals, more likely reflecting a true synchronous cross peak, while a large F value associated with a synchronous cross peak is more likely a false cross peak. Similarly, the noncoherence of asynchronous cross peaks can be checked by computing their F values. In this case, large F values are associated with true asynchronous correlation peaks, while small F values are more likely false asynchronous cross peaks. This approach has been tested on smoothed and unsmoothed near-infrared (NIR) spectra of human serum albumin (HSA) in aqueous solutions. A set of NIR spectra of HSA with increasing concentration were measured to generate 2D correlation maps. Unsmoothed spectra permitted the acquisition of synchronous and asynchronous 2D correlation maps containing all putative cross peaks in addition to artifacts generating by the fluctuation of noise or band overlapping. The control of coherence of pairs of signals on each cross peak, based on their F values, facilitated the interpretation of 2D-NIR maps, by improving the selection of true cross peaks from false cross peaks. The smoothed NIR spectra provided cross peaks without any artifacts in the synchronous and asynchronous correlation maps. The selection of cross peaks based on their F values confirmed the absence of false cross peaks in the case of smoothed spectra. Despite some limitations as discussed in this report, this procedure is straightforward to implement and will make the analysis of 2D vibrational spectra much easier.