Differentiation of human blood from animal blood using Raman spectroscopy: A survey of forensically relevant species

•Human and animal blood differentiation using Raman spectroscopy and chemometrics.•Incorporation of new, forensically relevant, animal species to expand applicability.•Validation of PLSDA methodology as highly selective for human and animal blood.•Differentiation of human and animal blood possible for unknown species of animals.•ROC analysis indicated 99% accuracy of PLSDA to differentiate human and animal blood. The identification of blood samples is a crucial facet of forensic investigations, particularly for violent crimes. One step in forensic serology (i.e., the analysis of bodily fluids) that is often skipped or overlooked is the determination that a bloodstain is of human or nonhuman origin. Typically, subsequent to identifying a stain as blood using a presumptive blood test, which have the propensity of providing false positive results, the stain is submitted for extraction of a DNA profile to compare with those in a database. It is extremely uncommon that evidentiary bloodstains are confirmed as being of human origin throughout the serological analysis. Therefore, time, money, and other resources can be wasted on obtaining a DNA profile from a bloodstain that may not be of human origin; if the intent was to obtain a human DNA profile and not that of an animal. This work demonstrates an important advancement of a previous study for nondestructive differentiation of human and animal blood using Raman spectroscopy coupled with partial least squares discriminant analysis (PLSDA). Raman spectra of blood from six species of animals, not previously accounted for, including chimpanzee, deer, elk, ferret, fish, and macaque were used to test a PLSDA classification method. These animal species are forensically relevant since they are (i) involved in wildlife crimes, (ii) consumed by humans, or (iii) known to produce a false positive result when their blood is tested with certain presumptive human blood tests. An external validation sensitivity of 1.00 and specificity of 0.93 for human class predictions was obtained from the PLSDA model constructed for this study. Using receiver operating characteristic (ROC) analysis of external human class predictions, the PLSDA model demonstrated 99% accuracy in being able to correctly classify any random blood sample as human or nonhuman. This is a significant advancement over the previous work and a very important finding as it demonstrates the superb selectivity of the developed method with high accuracy in being able t