Spectroscopy goes viral: Diagnosis of hepatitis B and C virus infection from human sera using ATR-FTIR spectroscopy

The development of a new fast, portable and reagent-free diagnostic technique for hepatitis B (HBV) and hepatitis C (HCV) viruses would be an enormous benefit to society. Here, we evalulate the ability of Attenuated Total Reflection Fourier Transform Infrared (ATR-FTIR) spectroscopy combined with multivariate data analysis to classify human serum samples based on the presence of HBV and HCV infection. Sera samples were prepared using three different methodologies: i) Sera depsoited onto glass cover slips, airdried and placed onto the ATR crystal. ii) Whole serum dried directly onto the ATR crystal. iii) Serum separated into high and low molecular weight compounds using a filtration approach and the high molecular weight fraction placed directly onto the ATR-FTIR diamond window and dried. For methodology i) the Partial Least Squares Discriminate Analysis (PLS-DA) calibration set included 313 (70 %) samples and the validation set 93 (30 %) samples. For HBV vs control the sensitivity and specificity was found to be 69.4 % and 73.7 % (10 latent variables (LV)), respectively. For HCV vs control the sensitivity and specificity was 51.3 % and 90.9 % (LV 11), respectively. In the second set of experiments the serum samples were dried directly onto the ATR diamond. PLS-DA models were constructed using 144 (70 %) samples for the calibration set and tested using an independent test set containing 62 (30 %) samples. For HBV versus control the sensitivity and the specificity was 84.4 % and 93.1 %, respectively (LV 8). For HCV versus control the sensitivity and specificity was 80.0 % and 97.2 %, respectively (LV 9). For HBV versus HCV the sensitivity and the specificity was 77.4 % and 83.3 %, respectively (LV 5). To increase the sensitivity and specificity serum sample was fractionated into high and low molecular weight components. In PLS-DA cross validated model (LV 8) the sensitivity and specificity was 87.5 % and 94.9 %, respectively for HBV vs control (high molecular concentrate). The PLS-DA cross-validated model (LV 8) for HCV vs control high molecular fraction produced a sensitivity and specificity of 81.6 % and 89.6 %, respectively. No linear correlation was observed for sera samples spiked with known viral loads using Partial Least Squares Regression (PLS-R) modelling.Spectra of positive serum (HBV and HCV) showed a strong band observed at 1631 cm−1, which was absent in the spectra of controls and assigned to the β-pleated sheet protein marker of immunoglobulin (Ig). A band at 1093 cm−1, observed in spectra of HBV infected sera, was assigned to CC and COmodes of polysaccharide N-glycan from hepatitis B surface antigen (HBsAg). The assignment was confirmed by atomic force microsocpy infrared (AFM-IR) spectroscopy of the isolated protein. This band represents a unique marker for HBV infection. In summary, ATR-FTIR spectroscopy is a powerful tool to study blood composition and identify potential disease markers but care must be taken to ensure that the modelling is not biased by inflammation markers, which may confound diagnosis