Method for forensic medicine identification of corpses in water based on metagenomics markers and application

Abstract

The invention belongs to the technical field of forensic medicine and biological medicine, discloses a method for forensic medicine identification of corpses in water based on metagenomics markers and application, and aims to solve the problems that existing drowning diagnosis is insufficient in accuracy and the deduction of post-death submerging time (PMSI) is limited. The method comprises the following steps: collecting a corpse lung tissue in water and a water sample in a corresponding water area, and obtaining a microbial DNA sequence by adopting a metagenome sequencing technology; the method comprises the following steps: processing data through bioinformatics analysis (such as fastp quality control, host sequence removal by KneadData and Kraken2 species annotation), and selecting and screening species level microbial markers in combination with a random forest algorithm and Boruta characteristics; and respectively establishing a drowning diagnosis model and a PMSI inference model based on the screened markers, and verifying the performance of the models through indexes such as AUC and MAE. Complete microbial communities of bacteria, eukaryotes (including fungi), archaea and viruses in lung tissues of corpses in water are comprehensively analyzed for the first time, 17 bacterial markers and 9 eukaryote markers are screened out for drowning diagnosis (verification experiment bacterial model AUC = 1, the accuracy rate is 89.29%, eukaryote model AUC = 0.95, the accuracy rate is 87.5%), and 17 markers are screened out for PMSI inference (integrated model MAE = 0.66 days). According to the method, the species resolution reaches the species level, the method is not influenced by amplification bias, the method is suitable for different water area environments, and an efficient and objective forensic medicine tool is provided for identifying the cause of the dead body in water and deducing the submerged time after the dead body.

Description

Technical Field

[0001] This invention belongs to the fields of forensic medicine and biomedicine, and relates to the application of metagenomic screening-based microbial biomarkers in determining whether a body in water has drowned and inferring the time of submersion after death. Background Technology

[0002] In-water body identification is a crucial part of forensic work. Besides determining whether the deceased drowned, it's essential to accurately estimate the postmortem submersion interval (PMSI), which is of significant forensic value in determining the nature of the case, judging the time of occurrence, and assisting in crime scene reconstruction. Currently, diatom testing is used in forensic practice to aid in determining drowning. However, the presence of false positives and false negatives in diatom testing limits its application in in-water body forensic identification. Compared to terrestrial bodies, research on postmortem submersion intervals is relatively limited both domestically and internationally. Some studies have proposed using a combination of total decomposition score (TADS) and accumulated degree day (ADD) to estimate the PMSI of in-water bodies, achieving good results. However, other studies suggest that this method is susceptible to subjective influences and is no longer applicable to early-stage in-water bodies. Therefore, forensic practice urgently needs to explore new testing methods and indicators with higher specificity and sensitivity to assist in identifying the cause of death and estimating the postmortem submersion interval in in-water bodies.

[0003] Microorganisms are a collective term for tiny organisms that are difficult to observe directly with the naked eye. They include cellular microorganisms such as bacteria, archaea, fungi, actinomycetes, protozoa, and algae, as well as organisms without complete cellular structures such as viruses, mycoplasma, and chlamydia. Microorganisms are characterized by their small size, rapid growth and reproduction, strong adaptability, and wide distribution, demonstrating great potential for practical application in solving problems such as postmortem interval (PMI) estimation and individual identification in forensic medicine. During drowning, a large amount of drowning fluid, such as freshwater or seawater, is inhaled into the respiratory tract and alveoli, leading to hypoxia and carbon dioxide retention, resulting in asphyxiation. Simultaneously, components of the drowning fluid, such as microorganisms, can enter the bloodstream through ruptured alveolar capillaries and spread throughout the body. Therefore, analyzing and exploring the succession patterns of microorganisms in aquatic corpse tissues / organs has become a new direction for solving drowning diagnosis and PMI estimation. High-throughput sequencing technology provides a powerful tool for detecting microorganisms. Currently, aquatic cadaver microbiology research primarily utilizes marker gene detection (e.g., 16S rDNA), which is efficient, low-cost, and suitable for analyzing samples with low biomass or a high proportion of host DNA. However, a single marker gene cannot simultaneously reflect information on multiple species in a sample, including bacteria, eukaryotes, and viruses. It carries limited data and has low resolution, making it difficult to delve into more specific species and strain taxonomic levels. Furthermore, primer and hypervariable region selection, as well as PCR amplification bias, significantly impact the detection results. Single marker gene detection and analysis are insufficient to meet the needs of scientific research and forensic practice. Metagenomics, also known as metagenomics, studies the entire microbial community in a specific habitat. Based on high-throughput sequencing technology, it obtains the sum of environmental microbial genome information, which is used to study the community structure, species classification, phylogenetics, gene function, and metabolic pathways of environmental microorganisms. Compared to marker gene detection, metagenomics can detect and analyze all DNA present in a sample, including eukaryotic DNA and viruses. Furthermore, metagenomic analysis has high resolution, enabling more precise classification and annotation of microorganisms, even down to the strain level. It can also directly obtain the relative abundance of functional genes and is less susceptible to amplification bias, demonstrating a clear advantage in microbial community analysis. This suggests that the discovery of new microbial indicators with high specificity and sensitivity based on metagenomics may provide a new and effective means to determine whether a body in water has drowned and to infer the time of submersion after death.

[0004] In summary, forensic identification of bodies in water lacks indicators with high specificity and sensitivity. Screening characteristic microbial indicators based on metagenomic data is expected to provide new evidence for the diagnosis of drowning and the estimation of submersion time in bodies in water. Summary of the Invention

[0005] To address the problems existing in current technologies, this invention provides a method and its application for determining whether a body in water has drowned and estimating the time of submersion based on metagenomic biomarkers. The method utilizes metagenomic detection of underwater body samples, combined with bioinformatics analysis and artificial intelligence algorithms, demonstrating the feasibility of metagenomic technology in forensic identification of bodies in water. Several microbial indicators (species level) are selected to establish a mathematical model for rapid and accurate diagnosis of drowning and estimation of the time of submersion.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solutions.

[0007] This invention discloses the application of a bacterial detection reagent in the preparation of a product for determining whether a body in water has died from drowning. The product is characterized by a combination of 17 bacterial markers in the test sample; the combination of the 17 bacteria is as follows: Aeromonas of animals (Aeromonas hydrophila) Leclercia adecarboxylata (Non-decarboxylated Leclercella) Citrobacter portucalensis (Citrobacter lucis) Citrobacter braakii (Citrobacter bromide) Citrobacter amalonaticus (Citrobacter malonic acid-free) Aeromonas rivipollensis (Aeromonas ripoile) Citrobacter pasteurii (Citrobacter pastoris) Aeromonas dhakensis (Aeromonas dacca) Citrobacter arsenates (Citrobacter arsenic) Aeromonas enchelia (Aeromonas elegans) Aeromonas hydrophila (Aeromonas hydrophila) Citrobacter freundii (Citrobacter freundii) Aeromonas salmonicida (Aeromonas salmonii) Aeromonas jandaei (Aeromonas japonicus) Aeromonas guinea pig (Aeromonas vulgaris) Micrococcus lylae (Lyra Micrococcus) and Aeromonas media (Aeromonas intermedia).

[0008] This invention also discloses the application of eukaryotic detection reagents in the preparation of products for determining whether a body in water has drowned, characterized in that the product detects a combination of nine eukaryotic biomarkers in the sample; the combination of the nine eukaryotic biomarkers is as follows: Brettanomyces bruxellensis (Brussels wine yeast) Virens stilaginoidea (Rice blast fungus) Talaromyces rugulosus (Fragaria foldii) Aspergillus chevalieri (Aspergillus chevaleri) Aspergillus oryzae (Aspergillus oryzae) Fusarium verticillium (Fusarium oxysporum) Dictyostelium discoideum (Discophyllus spp.) Aspergillus fumigatus (Aspergillus fumigatus) and Trypanosoma brucei (Treatisea bryony).

[0009] This invention also discloses the application of a bacterial detection reagent in the preparation of a product for estimating post-mortem submersion time, characterized in that the product detects a combination of 17 bacterial markers in the sample; the combination of the 17 bacteria is as follows: Hungatella xylanolytica (Hengartella xylan) Aerodenitrifying microorganisms (Aero-denitrifying Microbacterium) Selenobaculum gbiensis (Goblin) Bacteroides coprosuis (Bacteroides suis) Lacrimispora sphenoides (Pseudo-wedge-shaped teardrop-shaped sporogen) Acetoanaerobic sticklandii (Acid strychnine anaerobic bacteria) Clostridium sporogenes (Clostridium sporogenes) Aeromonas media (Aeromonas intermedius) Bacteroides thetaiotaomicron (Bacteroides polymorpha) Clostridium cochlearium (Clostridium sp.) Aeromonas guinea pig (Aeromonas vulgaris) Clostridium saccharobutylicum (Clostridium butyricum) Clostridium taeniosporum (Clostridium zoster) Clostridium botulinum (Clostridium botulinum) Aeromonas allosaccharophila (Aeromonas aberrantis) Clostridium cellulovorans (Clostridium cellulose) Bacteroides nordii (Bacteroides nordii).

[0010] This invention also discloses a method for determining whether a body in water has drowned based on metagenomic biomarkers, characterized by the following steps: (a) Obtain lung tissue samples from a human body in water; (b) Perform metagenomic sequencing on the samples to obtain microbial community sequence data of bacteria, eukaryotes, archaea and viruses; (c) Screening microbial biomarkers using the random forest algorithm and Boruta feature selection; (d) Construct a drowning diagnosis classification model using the screened microbial biomarkers.

[0011] Furthermore, the bacterial biomarkers for the drowning diagnostic model consist of the following 17 species: Aeromonas bestiarum, Leclercia adecarboxylata, Citrobacter portucalensis, Citrobacter braakii, Citrobacter amalonaticus, Aeromonas rivipollensis, Citrobacter pasteurii, Aeromonas dhakensis, Citrobacter arsenatis, Aeromonas encheleia, Aeromonas hydrophila, Citrobacter freundii, Aeromonas salmonicida, Aeromonas jandaei, Aeromonas caviae, Micrococcus lylae, Aeromonas media .

[0012] Furthermore, the eukaryotic biomarkers of the drowning diagnostic model consist of the following nine species: Brettanomyces bruxellensis, Ustilaginoidea virens, Talaromyces rugulosus, Aspergillus chevalieri, Aspergillus oryzae, Fusarium verticillioides, Dictyostelium discoideum, Aspergillus fumigatus, Trypanosoma brucei .

[0013] The present invention also discloses the application of any of the above-described methods in forensic identification products for drowned corpses and post-mortem bodies.

[0014] This invention also discloses a method for inferring post-mortem submersion time (PMSI) of a corpse in water based on metagenomic biomarkers, characterized by comprising the following steps: (1) Sample acquisition: collect lung tissue samples from the bodies in the water and water samples from the corresponding water areas; (2) Metagenomic sequencing: Microbial DNA was extracted from the sample, a sequencing library was constructed, and metagenomic sequencing was performed based on a high-throughput sequencing platform to obtain raw sequencing data; (3) Data preprocessing: The raw sequencing data were subjected to quality control, host sequence and species annotation were removed to obtain species-level microbial relative abundance data; (4) Screening of PMSI-related biomarkers: The random forest algorithm combined with the Boruta feature selection algorithm was used to screen biomarkers that change with PMSI from the microbial relative abundance data; (5) PMSI inference model establishment and verification: Based on the selected markers, a random forest regression model is constructed, and the model performance is verified by the mean absolute error (MAE) to realize PMSI inference.

[0015] Further, in step (4), the PMSI-related markers are: Hungatella xylanolytica, Microvirgula aerodenitrificans, Selenobaculum gbiensis, Bacteroides coprosuis, Lacrimispora sphenoides, Acetoanaerobium sticklandii, Clostridium sporogenes, Aeromonas media, Bacteroides thetaiotaomicron, Clostridium cochlearium, Aeromonas caviae, Clostridium saccharobutylicum, Clostridium taeniosporum, Clostridium botulinum, Aeromonas allosaccharophila, Clostridium cellulovorans, Bacteroides nordii .

[0016] This invention also discloses the application of any of the methods described above in forensic identification products for estimating the time of death of a body in water. Compared with the prior art, the beneficial effects of the present invention are as follows.

[0017] This invention provides the first comprehensive and systematic analysis of the microbial community (including bacteria, eukaryotes, archaea, and viruses) associated with corpses in water. Compared to existing amplicon sequencing (such as 16S rDNA and 18S rDNA), it contains richer microbial information and has higher species annotation resolution. The drowning diagnosis and post-mortem submersion time estimation model constructed based on metagenomic biomarkers has higher accuracy. Attached Figure Description

[0018] Figure 1 This is an overview of the microbial community composition at the boundary level for lung and water samples. Here, 'a' represents the species composition of the lung sample; and 'b' represents the species composition of the water sample.

[0019] Figure 2 These represent the genus-level species composition of lung microbial communities in different groups. Specifically, a represents the archaeal community composition in river water cadaver lung samples; b represents the bacterial community composition; c represents the eukaryotic community composition; d represents the viral community composition; e represents the archaeal community composition; f represents the bacterial community composition; g represents the eukaryotic community composition; and h represents the viral community composition.

[0020] Figure 3 These represent the genus-level species composition of microbial communities in different water bodies. Among them, a represents the archaea community composition in the water sample; b represents the bacterial community composition in the water sample; c represents the eukaryotic community composition in the water sample; and d represents the viral community composition in the water sample.

[0021] Figure 4 These represent the species composition at the species level of different groups of lung microbial communities. Specifically, a represents the archaic community composition in river water cadaver lung samples; b represents the bacterial community composition; c represents the eukaryotic community composition; d represents the viral community composition; e represents the archaic community composition; f represents the bacterial community composition; g represents the eukaryotic community composition; and h represents the viral community composition.

[0022] Figure 5 These represent the species composition of microbial communities at the species level in different aquatic environments. Among them, a represents the archaea community composition in the water sample; b represents the bacterial community composition in the water sample; c represents the eukaryotic community composition in the water sample; and d represents the viral community composition in the water sample.

[0023] Figure 6 These are the results of lung microbial community diversity analysis. Specifically, a represents the α-diversity of the archaea community in the lung samples; b represents the α-diversity of the bacterial community in the lung samples; c represents the α-diversity of the eukaryotic community in the lung samples; d represents the α-diversity of the viral community in the lung samples; e represents the β-diversity of the archaea community in the lung samples; f represents the β-diversity of the bacterial community in the lung samples; g represents the β-diversity of the eukaryotic community in the lung samples; and h represents the β-diversity of the viral community in the lung samples.

[0024] Figure 7These are the β-diversity analysis results of the microbial community in the lung samples from the river water group. Specifically, a represents the β-diversity of the archaic community in the lungs of corpses submerged in river water for 1 day; b represents the β-diversity of the bacterial community in the lungs of corpses submerged in river water for 1 day; c represents the β-diversity of the eukaryotic community in the lungs of corpses submerged in river water for 1 day; d represents the β-diversity of the viral community in the lungs of corpses submerged in river water for 1 day; e represents the β-diversity of the archaic community in the lungs of corpses submerged in river water for 3 days; f represents the β-diversity of the bacterial community in the lungs of corpses submerged in river water for 3 days; g represents the β-diversity of the eukaryotic community in the lungs of corpses submerged in river water for 3 days; h represents the β-diversity of the viral community in the lungs of corpses submerged in river water for 3 days; i represents the β-diversity of the archaic community in the lungs of corpses submerged in river water for 5 days; j represents the β-diversity of the bacterial community in the lungs of corpses submerged in river water for 5 days; k represents the β-diversity of the eukaryotic community in the lungs of corpses submerged in river water for 5 days; and l represents the β-diversity of the viral community in the lungs of corpses submerged in river water for 5 days.

[0025] Figure 8 These are the results of β-diversity analysis of the microbial community in lung samples from the estuary group. Specifically, a represents the β-diversity of the archaea community in the lungs of corpses submerged for 1 day at the estuary; b represents the β-diversity of the bacterial community in the lungs of corpses submerged for 1 day at the estuary; c represents the β-diversity of the eukaryotic community in the lungs of corpses submerged for 1 day at the estuary; d represents the β-diversity of the viral community in the lungs of corpses submerged for 1 day at the estuary; e represents the β-diversity of the archaea community in the lungs of corpses submerged for 3 days at the estuary; f represents the β-diversity of the bacterial community in the lungs of corpses submerged for 3 days at the estuary; g represents the β-diversity of the eukaryotic community in the lungs of corpses submerged for 3 days at the estuary; and h represents the β-diversity of the viral community in the lungs of corpses submerged for 3 days at the estuary. i represents the β diversity of archaeal communities in the lungs of corpses submerged for 5 days at the estuary; j represents the β diversity of bacterial communities in the lungs of corpses submerged for 5 days at the estuary; k represents the β diversity of eukaryotic communities in the lungs of corpses submerged for 5 days at the estuary; l represents the β diversity of viral communities in the lungs of corpses submerged for 5 days at the estuary; m represents the β diversity of archaeal communities in the lungs of corpses submerged for 10 days at the estuary; n represents the β diversity of bacterial communities in the lungs of corpses submerged for 10 days at the estuary; o represents the β diversity of eukaryotic communities in the lungs of corpses submerged for 10 days at the estuary; p represents the β diversity of viral communities in the lungs of corpses submerged for 10 days at the estuary.

[0026] Figure 9 This paper presents the establishment and validation of a comprehensive classification model for determining the cause of death based on bacterial communities. In the model diagram, a represents the multidimensional scaling (MDS) plot; b represents the ROC curves of the model's predictions for the exploratory and validation experimental samples; and c represents the confusion matrix of the model's predictions for the validation experimental samples.

[0027] Figure 10This study optimizes a bacterial community-based cause-of-death classification model using random forest and the Boruta algorithm. In the table, a represents the key species (bacteria) selected by the Boruta algorithm; b represents the cross-validation results of the random forest model; c represents the MDS plot of the optimized bacterial community-based cause-of-death classification model; d represents the ROC curves of the optimized model's predictions for exploratory and validation experimental samples; and e represents the confusion matrix of the optimized model's predictions for the validation experimental samples.

[0028] Figure 11 It is the relative abundance of potential biomarkers (bacteria) used to determine the cause of death.

[0029] Figure 12 This document presents the establishment, optimization, and validation results of a mortality classification model based on eukaryotic communities. Specifically, a) is the model's MDS diagram; b) is the ROC curve of the model's predictions for the exploratory and validation experimental samples; c) is the confusion matrix of the model's predictions for the validation experimental samples; d) is the important species (eukaryotes) selected by the Boruta algorithm; e) is the cross-validation result of the random forest model; f) is the MDS diagram of the optimized mortality classification model based on eukaryotic communities; g) is the ROC curve of the optimized mortality classification model based on eukaryotic communities for the exploratory and validation experimental samples; and h) is the confusion matrix of the optimized mortality classification model based on eukaryotic communities for the validation experimental samples.

[0030] Figure 13 It is the relative abundance of potential biomarkers (eukaryotes) used to determine the cause of death.

[0031] Figure 14 The results show the establishment, optimization, and validation of the PMSI inference model based on the lung microbiome. Specifically, a) represents the prediction results of the complete PMSI inference model based on the lung microbiome on the validation experimental samples; b) represents the cross-validation results of the random forest model; c) shows the variation of the relative abundance of the selected biomarkers with the extension of postmortem inundation time (PMSI); and d) represents the prediction results of the optimized PMSI inference model constructed based on the selected biomarkers on the validation experimental samples.

[0032] Figure 15 The Boruta algorithm is used to assess the importance of species to PMSI inference. Detailed Implementation

[0033] The present invention will be further described in detail below with reference to specific embodiments. However, this should not be construed as limiting the scope of the above-described subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0034] Unless otherwise specified, all reagents and materials used in this invention are commercially available.

[0035] Example

[0036] 1. Research subjects and grouping.

[0037] Exploratory Experiment: Ninety-six healthy adult male C57BL / 6J mice were randomly divided into a river group and a sea estuary group, with 48 mice in each group. The river group was conducted in the Dandong section of the Yalu River. The 48 mice were randomly divided into a drowning subgroup (D group) and a post-mortem immersion subgroup (PS group). Mice in the drowning subgroup were submerged in river water and subsequently immersed in the river. Mice in the post-mortem immersion subgroup were euthanized by cervical dislocation and then immersed in river water. The river water temperature was maintained at 20-25℃ during the experiment. At different time points post-mortem (1d, 3d, 5d, and 10d), six mice from each group were carcassed. The carcasses were dissected in a clean environment, and lung samples (approximately 100mg) were extracted using sterile instruments. After collection, the samples were immediately placed in liquid nitrogen and then stored at -80℃ for future analysis.

[0038] The estuary group was experimented at the mouth of the Yalu River, and the group design, animal model making and sample extraction were the same as those of the river water group.

[0039] Validation Experiment: Thirty-two healthy adult male C57BL / 6J mice were randomly divided into a river group and a sea estuary group, with 16 mice in each group. The river group was also conducted at the Dandong section of the Yalu River. The 16 mice were randomly divided into a drowning subgroup (D group) and a post-mortem immersion subgroup (PS group). Mice in the drowning subgroup were submerged in river water and subsequently immersed in the river. Mice in the post-mortem immersion subgroup were euthanized by cervical dislocation and then immersed in river water. The river water temperature was maintained at 20-25℃ during the experiment. At different time points post-mortem (1d, 3d, 5d, and 10d), two mice from each group were carcassed. The carcasses were dissected in a clean environment, and lung samples (approximately 100mg) were extracted using sterile instruments. After collection, the samples were immediately placed in liquid nitrogen and then stored at -80℃ for later analysis. The sea estuary group was conducted at the mouth of the Yalu River. The group design, animal model construction, and sample extraction were the same as in the river group of the validation experiment.

[0040] In addition, before the experiment began, water samples were extracted from the experimental water area, and a water sample filter membrane was made using a 0.2 μm pore size filter membrane and stored at -80℃ for future testing.

[0041] 2. Sample preparation and metagenomic sequencing.

[0042] Microbial DNA was extracted from samples using the QIAamp DNA Microbiome Kit (Qiagen, Hilden, Germany). Agarose gel electrophoresis (AGE) was used to analyze the purity and integrity of the DNA in each sample, ensuring that the amount of DNA used for library construction was ≥2µg and the concentration was >20 ng / µL. The DNA from qualified samples was randomly fragmented into approximately 350bp fragments using an ultrasonic disruptor. The entire library preparation process included end repair, A-tailing, adapter ligation, purification, and PCR amplification. Sequencing was performed using the Illumina NovaSeq X plus platform, generating 150bp paired-end sequencing results.

[0043] 3. Metagenomics data preprocessing.

[0044] FastP software was used for quality control, trimming low-quality (Q<30) bases at the ends of reads and filtering out sequences containing N, adapter contamination, or short sequences (<90 base pairs) to generate high-quality sequences. KneadData software was used to remove host sequences from the raw data, and FastQC software was used to evaluate the rationality and effectiveness of data quality control. Kraken2 software and the PlusPF database were used to perform taxonomic classification of the quality-controlled sequences into bacteria, archaea, eukaryotes, and viruses. For further analysis, species abundance was normalized using relative abundance.

[0045] 4. Metagenomics data analysis.

[0046] All statistical analyses were performed in the R environment (v4.1.1; http: / / www.r-project.org / ). To investigate the differences and variations in alpha diversity of microbial communities, the Chao1 index of taxonomic characteristics was estimated using the "diversity" function in the R package vegan. Distance analysis was performed using the adonis function in vegan. The proportion of variation in microbial community composition explained by cause of death (Group), site, or post-mortem inundation time (PMSI) was calculated using permutational multivariate analysis of variance (PERMANOVA), and visualized using principal coordinate analysis (PCoA). Analyses were performed separately for each microbial type.

[0047] To develop a classification model capable of distinguishing between samples with different causes of death (drowning and post-mortem submersion), a cause-of-death identification model was built based on species-level data using the "randomForest" software package (default parameters). Simultaneously, to infer post-mortem submersion time (PMSI), a PMSI inference model was built based on species-level data using the random forest algorithm with default parameters. The classification model was visualized by using multidimensional scaling (MDS) plots to show the similarity between samples and to assess the magnitude of differences between groups (i.e., different causes of death). The performance of the classification model was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC) and the accuracy of predictions on independent validation samples. The robustness of the regression model was evaluated using the mean absolute error (MAE) of the predictions on validation samples.

[0048] We employed the `rfcv()` function built into the `randomForest` package and the Boruta feature selection algorithm to screen microorganisms that could serve as biomarkers for drowning diagnosis and PMSI inference. In short, a nested cross-validation step was used to evaluate the relationship between the number of species and the model's prediction error rate. Building on our previous research, we limited the biomarker set to no more than 20 species. Based on the exploratory experimental dataset, we generated species importance scores using the Boruta algorithm. The Boruta algorithm categorizes species into three classes: confirmed, provisional, and excluded. Based on the species classified as "confirmed," we further screened for a target number of biomarkers.

[0049] Due to the breakage of the restraint cable, 10 days of samples were lost at the river water experimental site. 112 lung samples and 10 water samples were obtained during the exploration and validation experiments. To characterize the entire microbiome, shotgun metagenomic sequencing was performed on the lung and water samples. More than 35 million high-quality sequences were detected in each sample. First, the microbial community composition in the lungs and water samples was analyzed at different taxonomic levels. Bacteria constituted the largest proportion of the microbial community, followed by eukaryotes, while archaea and viruses were relatively less abundant. Figure 1 Subsequently, the relative composition of microbial species in the four kingdoms was compared at the genus level. Figure 2 , Figure 3 Significant differences in microbial community composition were observed between lung samples and water samples from both the river and estuary groups. Regarding archaea communities, most dominant genus groups were consistent across different water body samples, exhibiting similar patterns during putrefaction (e.g., ...). Nitrosopumilus ; Figure 2 a, e). In the bacterial community, with increasing submersion time, different samples from the two experimental water bodies, Aeromonas The abundance of all decreased. Furthermore, the abundance of [unclear] in drowning samples... Aeromonas The abundance was significantly higher than that of the post-mortem entry group. As the submersion time increased... Clostridium Gradually accumulating, especially in body samples found at the estuary ( Figure 2 b, f). In eukaryotic communities, the dominant genus of carcasses is consistent across different aquatic environments ( Figure 2 c, g), which contrasts with the pattern of viral communities ( Figure 2 d, h). Species-level analysis was performed using the same methods as at the genus level. Results showed significant differences in the species composition of the microbiome across the four kingdoms between lung and water samples. Figure 4 , Figure 5 High-abundance species in the lung microbiome exhibit spatial variability. During cadaver decomposition, key taxa (such as...) Nitrosopumilus piranensi and Proteus penneri The time-varying trend of ) remains consistent. It is worth noting that, including and Aeromonas veronii Indicator species, including those whose trans-aquatic variations are associated with causes of death.

[0050] The diversity of microbial community structure in lung samples was assessed using α and β diversity indices. The Chao1 index revealed that bacterial communities exhibited higher α-diversity compared to archaea, eukaryotes (including fungi), and viral communities. Figure 6 This result is consistent with the community composition observed at the boundary level. Within 10 days, the diversity of archaea, eukaryotes, and viruses changed very little over time, with no statistically significant differences between different causes of death. In stark contrast, there were significant differences in bacterial community diversity between the drowning and post-mortem water subgroups, particularly in the river water group. To visually assess the differences in β-diversity, principal coordinate analysis (PCoA) was performed on all samples based on the Bray-Curtis distance. Figure 6 eh). Gradual changes in microbial community structure over time were observed across all sampling intervals, with the most pronounced dynamic changes observed in the bacterial community. Notably, with increasing post-mortem submersion time (PMSI), samples consistently shifted towards the upper left quadrant on the PCoA plot. Figure 6 f). Community changes were significantly correlated not only with PMSI and cause of death, but also with experimental location. Among these factors, PMSI had the strongest impact on the structure of archaea, bacteria, eukaryotes, and viruses. While cause of death had a greater impact than location on most microbial assemblages, this pattern did not apply to viral communities. Overall, distinct microbial succession patterns were observed in all four microbial types in lung tissue. Compared to archaea, eukaryotes, and viruses, the bacterial component (lung bacterial community) was more sensitive to cause of death, location, and PMSI.

[0051] 5. Accurately classify drowning and post-mortem entry into water using lung bacteriology and eukaryotic characteristics.

[0052] Based on the above analysis, we found significant differences in the lung microbiota of cadavers from different locations (water bodies). To screen for drowning diagnostic microbial biomarkers applicable to cadavers in different water bodies and to establish a more universal model, we systematically compared the microbial community structure of drowned and post-mortem water samples over 10 days. According to the results of principal coordinate analysis (PCoA) and permutation multivariate analysis of variance (PERMANOVA), statistically significant differences were found in the bacterial communities of drowned and post-mortem water bodies from day 1 to day 10 of submersion, regardless of whether the body was in a river or at an estuary (p<0.01). Figure 7-8 Similarly, at both experimental sites, the eukaryotic communities in drowned and post-mortem water samples showed significant differences on days 3-5 (p<0.01). In contrast, archaea and viral communities did not exhibit a consistent pattern of differentiation: significant viral differences were only observed in samples from day 1 of river submersion, while archaea differences were observed at 5 days of river submersion (p<0.01) and 3 and 5 days of estuary submersion (p<0.05). These findings further demonstrate the significant differences in the lung microbiota of cadavers in water at different locations and also confirm the potential of bacteria and eukaryotes in the diagnosis of drowning.

[0053] Subsequently, we developed random forest classifiers for drowning diagnosis based on bacteria and eukaryotes, respectively. The MDS plot of the bacterial model (referred to as the complete bacterial classification model) showed a clear segregation trend among different causes of death. Figure 9 a), and has reliable diagnostic performance (exploratory experiment AUC=0.95, validation experiment AUC=1); Figure 9 b) Achieved 100% accuracy in external validation. Figure 9 c). Then, the Boruta algorithm was used to identify 56 distinguishable important bacterial species, among which... Aeromonas (13) Citrobacter (8) and Proteus (8) as the main ( Figure 10 a). Further cross-validation determined a minimal set of biomarkers consisting of 17 species ( Figure 10 (b, Table 1). The abundance of these markers differed significantly between the drowning and post-mortem entry subgroups in both experimental water locations (p<0.05). Figure 11 The diagnostic performance of these biomarkers was further validated using the random forest algorithm. An optimized bacterial classifier built based on the selected bacterial biomarkers showed stronger discriminative ability (AUC = 0.96 in the exploratory experiment and AUC = 1 in the validation experiment). Figure 10 c), with an accuracy rate of 89.29% ( Figure 10(d) This further demonstrates that these bacterial markers and the constructed model are applicable to bodies submerged for different periods in different water bodies. Furthermore, the same analysis was performed on eukaryotic communities in samples from 3-5 days ago, yielding a complete classification model. Samples from different causes of death showed a clear segregation trend ( Figure 12 a) The model has good diagnostic efficacy (AUC=0.90 in the exploratory experiment, AUC=0.91 in the validation experiment, and accuracy=87.5%). Figure 12 bc). Nine eukaryotic biomarkers were identified using the Boruta feature selection algorithm and cross-validation. Figure 12 (see Table 2). The abundance of these eukaryotic biomarkers differed significantly between the drowning group and the post-mortem water group in both experimental water locations. Figure 13 Finally, the simplified model built based on the screened eukaryotic biomarkers showed high accuracy (AUC = 0.94 in the exploratory experiment, AUC = 0.95 in the validation experiment, accuracy = 87.5%). Figure 12 (fh). Overall, our analysis highlights the reliable ability of bacterial and eukaryotic biomarkers to differentiate between ante-mortem drowning and post-mortem water entry, and their applicability to cadavers in various water bodies.

[0054] Table 1. Bacterial biomarkers used for identifying drowned and post-mortem bodies.

[0055] Table 2. Eukaryotic biomarkers used for identifying drowned and post-mortem bodies.

[0056] 6. Lung microbial community succession is an effective tool for inferring PMSI across aquatic environments.

[0057] Based on the above analysis, we found that different causes of death and different locations have a significant impact on the composition of the lung microbial community (especially the bacterial community) in aquatic corpses. To screen for more stable microbial biomarkers over time and thus establish a reliable PMSI inference model, we systematically evaluated the predictive ability of various microbial groups for PMSI inference. As shown in Table 3, the bacterial community exhibited excellent spatiotemporal stability across different causes of death and locations, explaining 83.85% of the community variance, with a mean absolute error (MAE) of 0.664 ± 0.088 days for the validation experimental samples, demonstrating high prediction accuracy. In comparison, the viral characteristic model showed slightly lower prediction accuracy (explaining 77.59% of the variance, with an MAE of 0.942 ± 0.187 days). The archaea model explained 75.12% of the variance, with an MAE of 1.214 ± 0.158 days; the eukaryotic model explained 75.28% of the variance, with an MAE of 1.413 ± 0.203 days, indicating relatively weaker inference abilities for both. Building upon the predictive capabilities of single microbial taxa, the fully integrated model incorporating cross-domain species significantly outperforms the single microbial taxa model. Figure 14 a). The integrated PMSI inference model based on multiple microbial types improved prediction accuracy (explained variance = 84.16%; MAE = 0.612 ± 0.084 days; Table 3, Figure 14 a) indicates a synergistic effect among different types of microorganisms. To further identify the key species for PMSI inference, we used the Boruta feature selection algorithm ( Figure 14 b, Figure 15 A total of 62 key predicted species were identified, of which 30 belong to Clostridium This highlights the significant contribution of this genus to predictive performance. Subsequently, based on the cross-validation results, we identified 17 potential biomarkers (Table 4), primarily bacteria. The abundance of these biomarkers gradually changed with increasing postmortem inundation time (PMSI). Figure 14 c). Finally, a random forest regression analysis was performed on the relative abundance of the 17 selected biomarkers to establish an optimized PMSI inference model. This model demonstrated satisfactory predictive ability for PMSI (mean absolute error = 0.66 ± 0.097 days). Figure 14 (d) indicates that it has robust performance across different causes and locations of death, providing a more widely applicable and reliable PMSI inference tool for forensic practice.

[0058] Table 3. Performance and error of postmortem inundation time estimation models constructed based on different types of microbial communities. Table 4. Biomarkers used for estimating post-mortem submersion time.

[0059] In summary, this invention comprehensively and systematically analyzes the composition and succession patterns of the entire microbial community of aquatic corpses using metagenomics technology, encompassing bacteria, eukaryotes, archaea, and viruses, providing a wealth of information for forensic identification of aquatic corpses. It improves the resolution of species annotation (species level), providing more accurate information for forensic identification of aquatic corpses. Through bioinformatics and artificial intelligence algorithms, key species (species level) suitable for forensic identification were screened, and accurate mathematical models were established, providing an efficient tool for forensic identification of aquatic corpses.

[0060] The above description is merely a preferred embodiment of the present invention and is not intended to limit the patent scope of the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. The application of bacterial detection reagents in the preparation of products for determining whether a body in water has drowned, characterized in that, The product test sample contained a combination of 17 bacterial markers; the 17 bacterial markers were as follows: Aeromonas bestiarum, Leclercia adecarboxylata, Citrobacter portucalensis, Citrobacter braakii, Citrobacter amalonaticus, Aeromonas rivipollensis, Citrobacter pasteurii, Aeromonas dhakensis, Citrobacter arsenates, Aeromonas encheleia, Aeromonas hydrophila, Citrobacter freundii, Aeromonas salmonicida, and Aeromonas... Aeromonas jandaei, Aeromonas caviae, Micrococcus lylae, and Aeromonas media.

2. The application of eukaryotic biological detection reagents in the preparation of products for determining whether a body in water has drowned, characterized in that, The product test sample contained a combination of nine eukaryotic biomarkers; the nine eukaryotic biomarkers were as follows: Brettanomyces bruxellensis, Ustilaginoidea virens, Talaromyces rugulosus, Aspergillus chevalieri, Aspergillus oryzae, Fusarium verticillioides, Dictyostelium discoideum, Aspergillus fumigatus, and Trypanosoma brucei.

3. The application of bacterial detection reagents in the preparation of products for determining post-mortem submersion time estimation, characterized in that, The product test sample contained a combination of 17 bacterial markers; the 17 bacterial markers were as follows: Hungatella xylanolytica, Microvirgula aerodenitrificans, Selenobaculum gbiensis, Bacteroides coprosuis, Lacrimispora sphenoides, Acetoanaerobium sticklandii, Clostridium sporogenes, Aeromonas media, Bacteroides thetaiotaomicron, Clostridium cochlearium, Aeromonas caviae, Clostridium saccharobutylicum, Clostridium taeniosporum, Clostridium... Clostridium botulinum, Aeromonas allosaccharophila, Clostridium cellulovorans, and Bacteroides nordii.

4. A method for determining whether a body in water has drowned based on metagenomic biomarkers, characterized in that... Includes the following steps: (a) Obtain lung tissue samples from a human body in water; (b) Perform metagenomic sequencing on the samples to obtain microbial community sequence data of bacteria, eukaryotes, archaea and viruses; (c) Screening microbial biomarkers using the random forest algorithm and Boruta feature selection; (d) Construct a drowning diagnosis classification model using the screened microbial biomarkers.

5. The method according to claim 4, characterized in that: The bacterial biomarkers for the drowning diagnostic model consist of the following 17 species. Composition: Aeromonas bestiarum, Leclercia adecarboxylata, Citrobacterportucalensis, Citrobacter braakii, Citrobacter amalonaticus, Aeromonasrivipollensis, Citrobacter pasteurii, Aeromonas dhakensis, Citrobacterarsenatis, Aeromonas encheleia, Aeromonas hydrophila, Citrobacter freundii, Aeromonas salmonicida, Aeromonas jandaei, Aeromonas caviae, Micrococcus lylae, Aeromonas media.

6. The method according to claim 4, characterized in that: The eukaryotic biomarkers for the drowning diagnostic model consist of the following nine species: Brettanomyces bruxellensis, Ustilaginoidea virens, Talaromyces rugulosus, Aspergillus chevalieri, Aspergillus oryzae, Fusarium verticillioides, Dictyostelium discoideum, Aspergillus fumigatus, and Trypanosoma brucei.

7. The application of the method described in any one of claims 4-6 in a forensic identification product for drowned corpses and post-mortem bodies.

8. A method for inferring post-mortem submersion time (PMSI) of a corpse in water based on metagenomic biomarkers, characterized in that, Includes the following steps: (1) Sample acquisition: collect lung tissue samples from the bodies in the water and water samples from the corresponding water areas; (2) Metagenomic sequencing: Microbial DNA was extracted from the sample, a sequencing library was constructed, and metagenomic sequencing was performed based on a high-throughput sequencing platform to obtain raw sequencing data; (3) Data preprocessing: The raw sequencing data were subjected to quality control, host sequence and species annotation were removed to obtain species-level microbial relative abundance data; (4) Screening of PMSI-related biomarkers: The random forest algorithm combined with the Boruta feature selection algorithm was used to screen biomarkers that change with PMSI from the microbial relative abundance data; (5) PMSI inference model establishment and verification: Based on the selected markers, a random forest regression model is constructed, and the model performance is verified by the mean absolute error (MAE) to realize PMSI inference.

9. The method according to claim 8, characterized in that, In step (4), the PMSI-related markers are Hungatella xylanolytica, Microvirgula aerodenitrificans, Selenobaculumbiensis, Bacteroides coprosuis, Lacrimispora sphenoides, Acetoanaerobiumsticklandii, Clostridium sporogenes, Aeromonas media, Bacteroidesthetaiotaomicron, Clostridium cochlearium, Aeromonas caviae, Clostridium saccharobutylicum, Clostridium taeniosporum, Clostridium botulinum, Aeromonasallosaccharophila, Clostridium cellulovorans, Bacteroides nordii.

10. The application of the method described in any one of claims 8-9 in a forensic identification product for estimating the time of death of a body in water.

Images