The discovery of body fluids during forensic investigation can provide countless clues to a crime. Saliva is often countered during certain investigations, particularly in the investigation of sexual assaults and other violent crimes. The wealth of chemical information housed in the complex matrix could provide clues as to who might have left the substance behind. One example of this is whether or not the donor is a smoker.

A team of researchers at the University at Albany have hit the news multiple times in recent years with their rapid, on-site tools for body fluid analysis. Using Raman spectroscopy, the group have successfully developed methods to identify body fluids, estimate body fluid age, and determine characteristics of the donor, such as race and sex. In the latest paper by Igor Lednev and his colleagues, published in Journal of Biophotonics, Raman spectroscopy has been shown to differentiate between smokers and non-smokers.

Working in collaboration with researchers at Kuwait University, the team applied Raman spectroscopy to dried saliva from smokers and non-smokers, aiming to use chemical differences in the samples to determine whether or not the donor smokes. Raman spectroscopy is a non-destructive technique that enables the rapid, on-site analysis of samples, producing distinctive chemical fingerprints consisting of bands produced by the interaction of light with molecular structures.

One might assume the test would target nicotine, a major chemical component in tobacco. However, nicotine is relatively short-lived in the body, thus is not a suitable target for analytical tests. Instead, the researchers focused on cotinine, a primary metabolite of nicotine with a notably longer half-life. Saliva samples from 32 donors were analysed by Raman spectroscopy and the chemical profiles produced studied for differences. Researchers soon encountered a problem. Raman bands indicative of cotinine overlapped with typical Raman bands produced by saliva, making the detection of cotinine in saliva challenging. The team used machine learning to solve this problem.

First, they identified eight spectral regions that contributed the most variation between the saliva of smokers and non-smokers. Using an artificial neural network, a classification model was constructed for the prediction of smoking habits of a donor. By inputting chemical data from known samples, the network is able to learn from the data in order to predict an output (in this case, whether or not the donor of a saliva sample was a smoker). In laboratory-based studies, the model constructed achieved an impressive accuracy of 100%.

Although this pilot study was based on a very limited sample size, the technique shows great promise in the determination of donor characteristics from dried body fluids.

Al-Hetlani et al. Differentiating smokers and nonsmokers based on Raman spectroscopy of oral fluid and advanced statistics for forensic applications. Journal of Biophotonics, 2019, DOI: 10.1002/jbio.201960123  https://onlinelibrary.wiley.com/doi/abs/10.1002/jbio.201960123