In mass disasters, armed conflicts, and criminal investigations, hard tissues such as bones and teeth are often the key materials for identification, particularly after soft tissue degradation.  Compared with conventional body fluid or soft tissue samples, DNA analysis of hard tissues is considerably more time- and resource-intensive, especially during sample preparation and extraction. Therefore, an effective pre-screening approach for hard tissue samples could reduce unnecessary resource expenditure and accelerate the recovery of informative DNA evidence. Previous studies have suggested that colorimetric analysis may be useful for predicting DNA quality in hard tissues. Building on this premise, the present study aimed to predict DNA quality through a more practical and convenient colorimetric analysis based on tooth photographs. Specifically, following image standardization with white-balance correction, multiple colorimetric features were extracted from 400 tooth images in the whole-tooth, crown, and root regions across several color spaces, namely RGB, L*a*b*, XYZ, and YCbCr. Spearman’s correlation analysis was then performed between these colorimetric variables and the number of detected alleles. The number of colorimetric variables significantly correlated with the number of detected alleles (P < 0.05) was highest in the root region, followed by the whole-tooth region, and lowest in the crown region. In the root region, the absolute correlation coefficients of significantly associated colorimetric variables ranged from 0.09 to 0.55, and the mean a* value showed the strongest correlation with the number of detected alleles. Based on these associations, we further evaluated their predictive utility for the number of detected alleles using machine learning models, including random forest, naive Bayes, logistic regression, support vector machine, and neural network approaches. The logistic regression model achieved the best performance, with a maximum accuracy of 86% on the test set. These findings suggest that tooth image colorimetry, combined with machine learning, may provide a practical and efficient strategy for the rapid pre-screening of forensic hard tissue samples. Using only images acquired by a smartphone or camera, this approach may assist in identifying samples with higher DNA testing value and thereby support more efficient allocation of forensic laboratory resources.

Authors

• Guihong Liu (Department of Forensic Genetics, West China School of Basic Medical Sciences and Forensic Medicine, Sichuan University, China)
• Shengqiu Qu (Department of Forensic Genetics, West China School of Basic Medical Sciences and Forensic Medicine, Sichuan University, China)
• Xiameng Chen (Department of Forensic Genetics, West China School of Basic Medical Sciences and Forensic Medicine, Sichuan University, China)
• Qiuyun Yang (Department of Forensic Genetics, West China School of Basic Medical Sciences and Forensic Medicine, Sichuan University, China)
• Peng Bai (Department of Forensic Genetics, West China School of Basic Medical Sciences and Forensic Medicine, Sichuan University, China)
• Weibo Liang (Department of Forensic Genetics, West China School of Basic Medical Sciences and Forensic Medicine, Sichuan University, China)
• Lin Zhang (Department of Forensic Genetics, West China School of Basic Medical Sciences and Forensic Medicine, Sichuan University, China)