Principal component analysis of terahertz spectrum on hemagglutinin protein and its antibody

Abstract

One has proved that the collective structural vibrational modes of proteins are in the terahertz (THz) frequency range. These frequencies relate to the polypeptide backbone and are thought to be essential for conformational dynamics necessary for protein function. Hemagglutinin (HA) is the main surface glycoprotein of the influenza A virus. The H9N2 subtype influenza A virus is recognized as the most possible pandemic strain as it crosses the species barrier, infects swine and humans. In this paper we use principal component analysis (PCA) to study the 7 different concentrations dependent terahertz spectra of hemagglutinin proteins, and detect the binding interaction of HA with the broadly neutralizing monoclonal antibody F10 in liquid phase. Spectrum pretreatment and band selection play a vital role in the THz spectroscopic analysis due to the fact that the original spectrum contains a large amount of interference information. In order to compress variables and extract useful information, we use a variety of pretreatment methods, such as second derivative, multiplicative scatter correction (MSC), least square polynomial fitting derivation, standard normalization, smoothing, moving window median filtering before PCA analysis. We even consider MSC + smoothing + SG second derivative + median filtering as the optimized pretreatment method finally. THz spectrum parameters including refractive index, absorption coefficient, reduced absorption cross-section and dielectric loss angle tangent are calculated in a frequency range of 0.1-1.4 THz for comparison. The results indicate that the reduced absorption cross-section presents the highest correlation response to the concentration variation of HA protein, and the dielectric loss angle tangent appears to be more appropriate for qualitative analysis of HA-antibody binding interaction. PCA method provides a feasible and effective way to find the sensitive parameters for further analyzing the function of protein and the antigen-antibody interaction using terahertz spectrum, whereas an appropriate pretreatment method is required.