Based on the data of 70 serial residential burglaries, this study examined how areas to commit crimes were developed across a series of crimes. Besides, methods for predicting future crime locations in offender profiling were considered. Size of convex polygon and nearest neighbor index (NNI) were calculated to examine how each offender’s offence space changed with the increase in their crime experience. Furthermore, three strategies proposed for predicting future crime locations in the context of geographical profiling, convex polygon, center of minimum distance (CMD), and kernel density estimation, were compared in terms of prediction precision. As the number of incidents increased, the size of convex polygon increased and NNI decreased, indicating that each offender’s whole crime area tended to spread out, but be more clustered in parallel across a series of crimes. For hit rates within the predicted area equal to the size of the convex polygon, kernel density estimation was most precise, following convex polygon and CMD. Prioritization of multiple clustered spots is essential for more precise prediction of future crime locations of an individual offender.

To enforce the law on cannabis use, it is necessary to establish a simple and efficient method for the extraction of cannabinoids and their metabolites from human biological specimens. Molecularly imprinted polymers (MIPs) are highly cross-linked polymers that can recognize and capture target compounds sterically and electrostatically. Solid-phase extraction (SPE) methods using MIP as adsorbents have been widely developed for selective extraction of target analytes. In this study, we evaluated the extraction efficacy of the following 10 cannabis-related compounds in human urine and plasma samples using a commercially available MIP-SPE cartridge: Δ⁹-tetrahydrocannabinol (THC), Δ⁸-THC, tetrahydrocannabivarin, cannabigerol, cannabichromene, cannabidiol, cannabinol, 11-nor-9-carboxy-Δ⁹-THC, 11-nor-9-carboxy-Δ⁸-THC, and 11-hybroxy-Δ⁹-THC. The MIP-SPE method combined with liquid chromatography–tandem mass spectrometry exhibited good linearity of calibration curves (correlation coefficients, > 0.994 for urine and > 0.999 for plasma) in the concentration ranges of 10–1000 ng/mL, detection sensitivities from sub-ng/mL to ng/mL levels, and satisfactory precisions (< 14 % for urine and < 4.8 % for plasma) and accuracies (within ± 21 % for urine and ± 11 % for plasma). The recovery rates were 49–102 % for urine and 34–87 % for plasma, and matrix effects were 49–91 % for urine and 45–108 % for plasma. The proposed method is expected to be applicable for the extraction and quantification of cannabinoids and their metabolites in actual human biological samples.

In this study, we evaluated likelihood ratio (LR) calculation methods for pairwise kinship analysis considering linkage and silent alleles. For the evaluation, we created a program that enables the calculation of LR for full siblings, half siblings, grandparent - grandchild, uncle/aunt - nephew/niece and first cousins, and compared calculated LRs with Familias and FamLink. We used the program to simulate LR distribution considering linkage and silent alleles in the Japanese population for the five relationships using GlobalFiler autosomal STRs. The simulation results indicated that the difference in the distribution of LR between full siblings and unrelated persons was large, suggesting high practicality of full sibling analysis by the LR calculation method in this study. Regarding full sibling analysis, we further evaluated the effect of silent alleles on LR and the cases when the two persons in fact have a relationship other than full siblings. Since the simulations of the pedigree data in this study were run under conditions in which mutation, linkage and silent alleles occur, the results should reflect reality better than the simulations in previous studies.

Protein toxins are toxic proteins produced by animals, plants, or microbes. Ricin and abrin are highly toxic protein toxins contained in plants seeds, castor beans (Ricinus communis) and rosary peas (Abrus precatorius), respectively. Ricin is also classified as a chemical warfare agent and restricted internationally by the Chemical Weapons Convention. Although those toxins with high purity are difficult to produce or obtain by ordinary people, the plants seeds are easy to obtain because those seeds are used in industry or used for gardening or accessories. Thus, murder attempts by contaminating those seeds into a beverage have occurred. To confirm the use of these seeds, the analysis of protein toxins themselves by biochemical methods are required. Nevertheless, those analyses are complicated and not easily implemented by scientists who are usually involved in the analysis of small molecules such as illicit drugs and poisons. On the other hand, small alkaloids ricinine and L-abrine are contained in castor beans and rosary peas, respectively. If analytical methods of those small molecules are used for the suspicious beverage sample initially, scientists can determine whether to proceed to the analysis of protein toxins or not. For biological samples, screening methods of ricin or abrin exposure by detecting ricinine or L-abrine have been reported by some groups. However, simultaneous screening method for beverage samples have not been reported so far. In this work, we have developed simultaneous LC-MS/MS method of ricinine and L-abrine using a phenyl column and applied the optimized method to beverage samples. After simple pretreatments involving deproteinization and solid phase extraction, ricinine and L-abrine could be detected at low concentration and the quantitation results were also sufficient for screening purpose. The method was also applied to beverage samples contaminated with castor beans, and applicability to real samples was confirmed.

In kinship analysis by DNA typing, genotypes of short tandem repeats of the deceased and relatives are tested, and the likelihood ratio is calculated to evaluate the kinship relationship. However, some alleles might not be detected (i.e., allele drop-out) in a small amount or degraded samples. For such samples, the usual likelihood ratio calculation cannot be performed. Therefore, in this study, we aimed to develop a likelihood ratio calculation software for allele drop-out. For this purpose, we updated KinBN software that we previously developed. Bayesian networks were used to consider mutation, linkage between loci, and allele drop-out. For loci with the possibility of allele drop-out, the likelihood ratio is calculated assuming both homozygous and heterozygous genotypes. In addition, various calculation conditions including probability of allele drop-out can be freely set by the user within the software. Using the updated KinBN, likelihood ratios were calculated for a small amount of samples in which some alleles were not detected, assuming a paternity or sibling test. As a result, the likelihood ratio could be calculated even when shared alleles were not detected among blood relatives due to allele drop-out. Furthermore, the overall likelihood ratios of GlobalFiler Kit showed high values for related individuals and low values for unrelated individuals, indicating that this method is useful for the identification of relatives. However, the risk of false positives was suggested to be increased in samples with multiple drop-out. KinBN is available on the website for anyone to use, and it is expected to be applied to future identification work such as disaster victim identification.

Hair samples are commonly utilized as evidence in criminal investigations. Typically, hair collected at crime scenes undergoes morphological examination with the naked eye and optical microscopes at forensic science laboratories, followed by DNA profiling. However, when human and animal hairs are intermixed at a crime scene, the volume of hair evidence becomes vast, necessitating considerable time and effort from collection to DNA analysis.

In this study, we explored the feasibility of using convolutional neural network models to screen for human hairs from samples collected at crime scenes. Initially, images of the root and shaft sections of cat and dog hairs, as well as human hairs, were captured using a portable digital microscope and smartphone. Images of the tip sections of human hairs were also taken, creating an image dataset comprising seven classes.

For these seven classes of image data, we developed image classification models by training a DenseNet-121 convolutional neural network from scratch or fine-tuning a pre-trained DenseNet-121 with ImageNet. The optimization functions employed were SGD or Adam, and data augmentation included basic horizontal flipping, rotation, and brightness adjustment, or more enhanced blurring, noise, and color distortion.

The results of the training revealed that the model fine-tuned with Adam and basic data augmentation achieved the highest accuracy of 99 %. However, Grad-CAM++ indicated that this model sometimes focused on the background rather than the hair within the images.

Conversely, the model fine-tuned with SGD and enhanced data augmentation exhibited a lower accuracy of 93.71 % compared to the aforementioned model but was the most reliable in focusing on the hair within the images and robust against images featuring only shapes. This model also demonstrated high precision and recall rates for human hair roots. These outcomes suggest that this model has the potential to screen human hairs with high accuracy.