In this study, Raman spectroscopy was combined with advanced statistical analysis and used for semen identification.
Proper identification of body fluids at crime scenes is imperative to forensic investigations, as they yield crucial pieces of DNA evidence that lead to subsequent investigation resolution and criminal justice. Nevertheless, body fluid identification can prove a burdensome process, since several ubiquitous substances found at crime scenes often resemble the appearance of a specific body fluid; therefore, it is extremely important to find a reliable method for body fluid discovery and differentiation among other potential environmental inferences (EIs). The method used in the current project has already been proven to be a powerful tool for characterization of body fluid traces for forensic purposes. Twenty-seven EI substances that may (i) yield a false positive (FP) result on routine forensic tests for semen and/or (ii) resemble a semen stain at a crime scene were examined. Raman spectra of EI substances were compared to spectra of main body fluids: semen, blood, saliva, sweat, and vaginal fluid. The random forest algorithm was utilized for differentiating between body fluids and EIs commonly found for semen traces. Using the classification probability threshold of 70 percent, entire separation of classes was observed. None of the 27 EI substances was classified as a body fluid. This confirms a great potential of Raman spectroscopy for non-destructive and rapid identification of semen traces found at a crime scene. (publisher abstract modified)