A Raman algorithm to estimate human age from protein structural variations in autopsy skin samples: a protein biological clock

Abstract

The recent increase of the number of unidentified cadavers has become a serious problem throughout the world. As a simple and objective method for age estimation, we attempted to utilize Raman spectrometry for forensic identification. Raman spectroscopy is an optical-based vibrational spectroscopic technique that provides detailed information regarding a sample's molecular composition and structures. Building upon our previous proof-of-concept study, we measured the Raman spectra of abdominal skin samples from 132 autopsy cases and the protein-folding intensity ratio, R_PF, defined as the ratio between the Raman signals from a random coil an α-helix. There was a strong negative correlation between age and R_PF with a Pearson correlation coefficient of r = 0.878. Four models, based on linear (R_PF), squared (R_PF²), sex, and R_PF by sex interaction terms, were examined. The results of cross validation suggested that the second model including linear and squared terms was the best model with the lowest root mean squared error (11.3 years of age) and the highest coefficient of determination (0.743). Our results indicate that the there was a high correlation between the age and R_PF and the Raman biological clock of protein folding can be used as a simple and objective forensic age estimation method for unidentified cadavers.

Figure 1

Explanatory schema of the different secondary structures involved with the spectroscopic measurement of human age based on the protein folding mechanism. Three kinds of protein mesostructures (shown in Table S2) are detected by Raman spectroscopy. Our previous study suggested that the rate of random coiled disordered protein structures (Raman signal seen at 1681 cm⁻¹) becoming α-helix structures (Raman signal at 1652 cm⁻¹) was age dependent.

Figure 2

Prediction curve using Model 2 (R_PF). Complete plots for all age intervals are given for the protein-folding intensity ratio (R_PF) and the 95% prediction intervals are also shown. There was a strong negative correlation between age and the R_PF with a Pearson correlation coefficient of r = 0.878.

Figure 3

Average Raman spectra from skin samples from the patients listed in Table S1 for the selected age intervals (shown in inset). Spectroscopic related to the proteins are shown. In the protein spectral zone, the labels α and d refer to α-helix and disordered structures, respectively. The vibrational origins of all labeled bands are listed in Table S2 of the Supplementary Materials and are visualized in Fig. 1.