An initial exploration of machine learning for establishing associations between genetic markers and THC levels in Cannabis sativa samples.

Cannabis sativa, a globally commercialized plant used for medicinal, food, fiber production, and recreation, necessitates effective identification to distinguish legal and illegal varieties in forensic contexts. This research utilizes multivariate statistical models and Machine Learning approaches to establish correlations between specific genotypes and tetrahydrocannabinol (Δ9-THC) content (%) in C. sativa samples. 132 cannabis leaves samples were obtained from legal growers in Piedmont, Italy, and illegal drug seizures in Turin. Samples were genetically profiled using a 13-loci STR multiplex and their Δ9-THC content was detected through quantitative GC-MS analysis. This study aims to assess the use of supervised classification modelling on genetic data to distinguish cannabis samples into legal and illegal categories, revealing distinct clusters characterized by unique allele profiles and THC content. t-distributed Stochastic Neighbor Embedding (t-SNE), Random Forest (RF) and Partial Least Squares Regression (PLS-R) were executed for the machine learning modelling. All the tested models resulted effective discriminating between legal samples and illegal. Although further validation is necessary, this study presents a novel forensic investigative approach, potentially aiding law enforcement in significant marijuana seizures or tracking illicit drug trafficking routes.

A comparison study of the detection of bloodstains on painted and cleaned surfaces with luminol.

There seems to be a limited amount of research about the detection of concealed bloodstains on painted surfaces. The bloodstains on walls and floors are often removed by cleaning, in some cases the surfaces are painted by the perpetrator after committing a violent crime in order to hide the crime that has occurred. The study hereafter extends and deepens on previous researches by investigating the detectability of horse bloodstains on painted ceramic tiles as a function of the number of layers of paint. In this study luminol was used as a reagent to detect the bloodstains. The study focuses on two types of paints: water based and solvent based paint. This study also investigates the effectiveness in reducing the detectability of bloodstains on ceramic tiles using four different cleaning methods pure water, soap with water, wet wipes, and bleach. In the experiment the bloodstains were cleaned at various intervals of time after the deposition (two minutes, fifteen minutes and one hour). The study concluded that the bloodstains concealed by layers of solvent based paint are less likely to be detected by luminol compared to water based paint. The study also concluded that the tiles cleaned with bleach are recognisable from the other ones cleaned using other methods. In each study the duration of the reaction was timed, highlighting the differences in the cleaning methods.