CN113341044B - A method for identifying drowning based on metabolomics markers and its application - Google Patents

Abstract

The invention belongs to the technical fields of forensic science and biomedicine, and in particular relates to the application of metabolites screened based on metabolomics in judging whether an early water corpse is drowned. The marker for judging whether an early water corpse is drowned is characterized in that the marker includes any one or a combination of more than one of the following 14 small molecule metabolites: lactic acid, mevalonic acid, Phenylacetylglycine, tetradecanedioic acid, inosine, malic acid, beta hydroxybutyrate, undecanoic acid, glycylleucine, corticosterone, thiamine, acetylcarnitine, alanylleucine acid and kynurenine. A method for identifying drowning based on metabolomics markers provided by the present invention has a higher predictive ability based on comprehensive analysis of a large number of biomarkers than the method using one or a few biomarkers. The content of endogenous target biomarkers, combined with machine learning algorithms, can easily and accurately identify fresh drowned corpses, which is expected to provide more favorable assistance for case detection in forensic practice.

Description

A method for identifying drowning based on metabolomics markers and its application

technical field

The invention belongs to the technical field of forensic science and biomedicine, and relates to the application of metabolites screened based on metabolomics in judging whether an early water corpse is drowned.

Background technique

Drowning is the suffocation death caused by the lack of oxygen and carbon dioxide retention in the body due to the obstruction of the respiratory tract and alveoli by the drowning liquid after the body falls into the water. According to the statistics of the World Health Organization, about 360,000 people die from drowning every year in the world, and drowning has become the third leading cause of accidental death among people. In the judicial practice of forensic medicine, there are frequent cases of killing victims by other means and throwing their corpses into the water to fake drowning. Therefore, it is very important to accurately identify whether the corpse in the water is drowned or not for further case investigation.

During the drowning process of the body, diatoms, chlorophyll, microorganisms, etc. in natural water can enter the blood circulation through damaged alveolar capillaries, and then spread to other organs. At present, in the practice of forensic medicine, diatom detection is used to assist in the judgment of drowning. The detection of diatoms has become the "gold standard" for drowning identification, that is, the detection of diatoms in multiple organs (such as lungs, liver, kidneys, etc.) The deceased inhaled a large amount of drowning fluid during his lifetime, which provided a reference for identification of drowning. But on the one hand, some scholars have detected diatoms in non-drowning corpses, which may be related to diatoms existing in the soil or even in the air. questioned. On the other hand, in the case of drowning in water with low or no diatom content, diatoms may not be detected in multiple organ specimens obtained using existing methods. Due to the existence of false positives and false negatives in the test results of diatoms, the application of diatoms in the diagnosis of drowning is limited to a certain extent. Therefore, it is urgent to find new testing methods with high specificity and sensitivity in forensic practice. And indicators to assist in identifying whether the corpse in the water is drowned.

In recent years, studies have shown that the body has experienced complex processes and mechanisms during drowning, including fear of drowning, diving responses, upper airway reflexes, swallowing, vomiting, drowning liquid inhaled into the alveoli and entering the blood with air and blood exchange, blood cell rupture, Electrolyte disorders, etc., and an in-depth understanding of the interaction of these mechanisms and changes in multiple metabolic pathways can help us understand the process of drowning, and then look for auxiliary diagnostic indicators for drowning with high specificity and sensitivity. With the improvement of the performance of corresponding instruments and the development and establishment of metabolomics methods, it is possible to analyze the changes in the metabolic process of the body in a short period of time from an overall perspective. Metabolomics is an emerging method of systems biology. By detecting the content of various small-molecular-weight metabolites in biological fluids or tissues and performing mathematical modeling and analysis, high-dimensional data with rich information can be obtained for the study of diseases and external The effects of stimuli on the organism provide new directions. By combining advanced analytical techniques (such as GC-MS, LC-MS, and NMR) with high-throughput bioinformatics tools, metabolomics has been attempted to solve scientific problems in the fields of clinical medicine and forensic science. Compared with traditional statistical analysis methods, machine learning methods are often better suited to solving classification problems involving a large number of latent indicators. Random forest is one of many machine learning methods. It uses a randomly selected subset of features to build multiple decision trees and combines these decision trees to form a forest to predict the outcome of a classification problem. By randomly selecting subsets and combining the analysis results of multiple decision trees, the errors caused by bias and variance can be minimized. At the same time, random forests show good generalization ability and robustness when applied to high-dimensional data sets. , suggesting that our random forest analysis method based on metabolomics high-throughput data may provide a new idea for judging whether a corpse in water is drowned.

In summary, there is still a lack of indicators with high specificity and sensitivity for judging whether an early water corpse is drowned. Screening characteristic indicators based on metabolomics data is expected to provide a new basis for judging whether a water corpse is drowned.

Contents of the invention

In view of the above problems, the present invention provides a method for identifying drowning based on metabolomics screening markers and its application. The present invention proves the feasibility of metabolomics technology in judging whether a corpse in fresh water is drowned, and applies the screened A number of small molecule metabolite indicators were used to establish mathematical models to quickly and accurately determine whether the early water corpses were drowned.

In order to achieve the above object, the present invention provides the following technical solutions.

The present invention provides a marker for judging whether an early corpse in water is drowned, characterized in that the marker includes any one or a combination of more than one of the following 14 small molecule metabolites: lactic acid, Mevalonate, Phenylacetylglycine, Tetradecanedioic Acid, Inosine, Malate, Beta Hydroxybutyrate, Undecanoic Acid, Glycylleucine, Corticosterone, Thiamine, Acetylcarnitine, Alanyl Leucine and Kynurenine.

The present invention also provides a screening method for the above-mentioned markers, and the specific steps of the screening method are as follows:

Step 1. Randomly divide the heart blood samples extracted from the dead bodies of the rats in the drowning group and the dead bodies into the water after death into a training set and a verification set;

Step 2, using ultra-high performance liquid chromatography tandem mass spectrometry UHPLC-MS/MS system to perform metabolomics detection, and obtain metabolic fingerprints of various small molecule metabolites;

Step 3. Import the training set data into the R (version 3.6.1) language, use the randomForest program package to take the sample grouping as the dependent variable and various metabolites as the independent variable, and randomly select the training set samples with replacement to establish a random The forest classifier model finally obtains the relative importance of independent variables and the prediction accuracy of the model. The relative importance of the independent variable is expressed by the average accuracy decline as an indicator. This value refers to the degree of reduction in the prediction accuracy of the random forest classifier model when the value of a variable is changed to a random number. The greater the value, the greater the importance of the variable. The greater the accuracy, the prediction accuracy of the model is represented by the area under the receiver operating characteristic curve and the accuracy rate of the model on the verification set. The accuracy rate is the ratio of the number of correctly classified verification samples to the number of all verification samples, AUC and accuracy rate The closer to 1, the more reliable the model is;

Step 4. Cross-check the random forest classifier model established with all metabolites as independent variables, use the upper limit of the 95% confidence interval of the cross-check error of the classifier model established in step 3 as a reference, and use the same cross-check error as a standard , to build a simple classifier model containing fewer metabolites;

Step 5. According to the relative importance of each metabolite in the random forest classifier model described in step 3 and the variable projection importance value in the PLS-DA model, select and determine 14 pairs of pairs for judging whether the corpse in fresh water is drowned marked metabolites.

The present invention also provides a method for constructing a simple classifier model for the above-mentioned 14 metabolic markers, and for judging whether a dead body in fresh water is drowned. It is characterized in that the specific steps of the construction method are as follows:

Step 1. Extract the content of 14 markers and the grouping data of the cause of death in each sample of the training set and import them into R. With the help of the randomForest package, the sample grouping is used as the dependent variable, and the 14 markers are used as the independent variable. Extract training set samples to establish a random forest simple classifier model;

Step 2. Input the content information of the corresponding 14 markers in the verification set samples into the constructed random forest simple classifier model, obtain the prediction results of the model for each sample group, and analyze them.

Compared with the prior art, the present invention has beneficial effects.

(1) The present invention establishes animal models in natural river water (fresh water) to simulate human corpses found in the water environment, and provides a method for identifying drowning based on metabolomics markers, which is conducive to the transformation of results.

(2) The present invention provides markers for judging whether an early corpse in water is drowned. The 14 markers are closely related to the cause of death of corpses in water, are relatively less affected by postmortem changes, and have differences among parallel samples. It has the characteristics of small size and high contribution to model prediction accuracy in the random forest machine learning algorithm.

(3) The present invention provides a method for identifying drowning based on metabolomics markers. Metabolomics technology can detect multiple metabolites at the same time. Microscopic changes during early postmortem degradation.

(4) A method for identifying drowning based on metabolomics markers provided by the present invention is based on a comprehensive analysis of a large number of biomarkers, which has a higher predictive ability than the method using one or a few biomarkers. The content of several endogenous target biomarkers in the sample, combined with machine learning algorithms, can conveniently and accurately identify fresh drowned corpses, which is expected to provide more favorable assistance for case detection in forensic practice.

Description of drawings

Figure 1 shows the observation results of metabolic characteristics in different groups of cadaver heart blood samples in the training set.

Figure 2 is the PLS-DA model built with training set samples.

Figure 3 shows the results of 200 permutation tests on the PLS-DA model.

Figure 4 is the ROC curve of the random forest classifier model built for all metabolites in the validation set.

Figure 5 is the result of the 5-fold 10-fold cross-validation of the initially established random forest classifier model.

Figure 6 is a box diagram of the content differences and changes of 14 markers among different causes of death in early water corpses.

Figure 7 is the ROC curve in the verification set of the random forest classifier model constructed based on the differences in the contents of 14 markers.

Detailed ways

The specific implementation manners of the present invention will be further described in detail below in conjunction with the accompanying drawings and examples. The following examples illustrate the invention in detail, but are not intended to limit the scope of the invention. This embodiment adopts conventional experimental techniques, which are familiar to those skilled in the art, and can be carried out according to the instructions provided by the material manufacturer according to this embodiment.

1. Research objects and groups.

Training set: 80 male SD rats were randomly divided into a drowning group (group D) and a postmortem submersion group (group PS), with 40 rats in each group. The rats in the drowning group were immersed in river water and drowned, and the rats in the water-entrance group were killed by asphyxiation with CO ₂ gas and then immersed in river water. The temperature of the river water was 20-25°C during the experiment. At different time points after death (0h, 6h, 12h, 18h and 24h), 8 dead rats were taken from each group, and heart blood (about 200 μL) was extracted. Save for future inspection.

Verification set: 20 male SD rats were randomly divided into a drowning group and a water submersion group after death, with 10 rats in each group. They were sacrificed according to the above method and immersed in river water. At 0h, 6h, 12h, 18h and 24h after death, 2 rats in each group were taken to extract heart blood, and then the samples were stored at -80°C for future inspection.

2. Metabolite extraction: Take 100 μL of heart blood sample and put it in a new EP tube, add 400 μL of mass spectrometry grade methanol to precipitate protein, vortex, let stand in ice bath for 5 min, centrifuge at 15000 g, 4°C for 10 min, and take 100 μL of supernatant The solution was diluted with mass spectrometry-grade water to a methanol content of 53%, and placed in a centrifuge tube at 15,000 g, centrifuged at 4°C for 10 min, and the supernatant was collected for analysis on the machine.

Take an equal volume sample from each experimental sample and mix it as a QC sample, which is used to balance the chromatography-mass spectrometry system and monitor the status of the instrument, and evaluate the system stability during the entire experiment. Take 53% methanol aqueous solution as the blank sample, and the treatment process is the same as that of the experimental sample.

3. Metabolic profile detection.

UHPLC-MS/MS system was used to detect the content of various small molecule metabolites in the samples.

LC-MS/MS analysis was performed using a Vanquish UHPLC system (Thermo Fisher) and an Orbitrap QExactive series mass spectrometer (Thermo Fisher).

(1) Instrument conditions.

Chromatographic conditions:

Chromatographic column: Hyperil Gold column (C18)

Column temperature: 40°C

Flow rate: 0.2 mL/min

Positive ion mode: mobile phase A: 0.1% formic acid

Mobile Phase B: Methanol

Negative ion mode: mobile phase A: 5mM ammonium acetate, pH 9.0

Mobile Phase B: Methanol.

The chromatographic gradient elution program is shown in Table 1.

Table 1. Chromatographic gradient elution program

。

.

Mass spectrometry conditions: scan range selection m/z 100-1500; ESI source settings are as follows: spray voltage: 3.2 kV; sheath gas flow rate: 35 arb; auxiliary gas flow rate: 10 arb; ion transfer tube temperature: 320 ℃; polarity: Positive ion mode; negative ion mode; MS/MS secondary scan is a data-dependent scan.

Metabolite identification: Import the off-machine data (.raw) file into Compound Discoverer 3.1 (CD) library search software, perform simple screening of parameters such as retention time and mass-to-charge ratio, and then analyze different samples according to the retention time deviation of 0.2min and mass Perform peak alignment with a deviation of 5ppm to make the identification more accurate, and then perform peak extraction according to the set mass deviation of 5ppm, signal intensity deviation of 30%, signal-to-noise ratio of 3, minimum signal intensity of 100000, and summation ions, etc., and quantify the peak area at the same time , and then integrate the target ions, then predict the molecular formula through the molecular ion peak and fragment ion and compare it with the mzCloud, mzVault and MassList databases, use the blank sample to remove the background ions, and normalize the quantitative results, and finally get various Qualitative and quantitative results of metabolites.

4. Differences in metabolic profiles of cadaver blood samples in fresh water.

After metabolic profile detection and data preprocessing, a total of 601 metabolites were detected. Firstly, Principal Component Analysis (PCA) was performed using all the identified metabolites to observe the metabolic characteristics in heart blood samples from cadavers with different causes of death.

The results are shown in Figure 1. The QC samples were well aggregated, indicating that the instrument was stable and the test results could be used for in-depth analysis. In the score chart, there is an obvious separation trend in each time group, indicating that the metabolic profiles of heart blood samples with different postmortem submersion intervals (Postmortem Submersion Interval, PMSI) have obvious differences. However, the effect of distinguishing between different causes of death is not ideal, which may be because the differences in metabolic levels between different causes of death are covered by the differences at different times. Further exploration of the differences in metabolic profiles between different causes of death is needed.

Then, a Partial Least Squares Discriminant Analysis (PLS-DA) model was established using the training set samples, as shown in Figure 2, to further distinguish different causes of death. The results showed that the two causes of death were clearly separated, and the model had a better explanation and predictability (R2Y=0.939, Q2Y=0.834), indicating that fresh water corpses with different causes of death have significant differences. In order to verify whether the model has overfitting phenomenon, the patent of the present invention has carried out permutation tests on the model (200 times), as shown in Figure 3, the result shows that the Q2 intercept is negative, and the PLS-DA model is not overfitted. At the same time, the VIP value of each metabolite was obtained, which can measure the impact strength and explanatory power of each metabolite difference on the classification and discrimination of samples in each group.

5. Establishment and validation of a discriminant model for freshly drowned corpses.

The above results show that there are always differences in metabolic profiles in the early postmortem period (within 24 hours) between drowned and postmortem corpses. It is conjectured that this difference can provide important reference information for judging whether the corpse in fresh water is drowned. Therefore, based on the content differences of various metabolites and combined with the random forest algorithm, a drowned corpse discrimination model was established, and the model was verified using the validation set samples. In the training set, drowning and post-mortem dumping are used as dependent variables, and various metabolites are used as independent variables, and the training set samples are randomly selected with replacement for random forest classification, and finally the error of the model and the relative importance of each variable are sorted . The error of the model is represented by Out-of-bag (OOB) error. OOB refers to the part of the data (about 1/3) that is not sampled when the training set data is randomly sampled with replacement. The OOB error refers to the percentage of the number of model misidentifications to the number of all out-of-bag data when the model constructed by drawing samples is verified in the out-of-bag data. The relative importance of metabolites was assessed by the decrease in the average accuracy of the model. In order to avoid the impact of different parameter settings of the model, the model was established and verified with default parameters. The results showed that the AUC of the model in the verification set was 1, and 19 samples out of 20 samples in the verification set could accurately identify the cause of death. The classifier model predicted accurately The rate can reach 95%, as shown in Figure 4 and Table 2. It shows that metabolomics can be used to identify drowned corpses, which has important reference value for forensic practice.

Table 2 Prediction results of the grouping of validation set samples by the cause of death discriminant model established with all metabolites

。

.

6. Biomarker screening, model establishment and verification for identifying drowned corpses.

A large amount of effective information has been obtained by analyzing the entire metabolic profile, and the feasibility of metabolomics in the identification of freshly drowned corpses has been proved. However, the study of the entire metabolic profile requires a lot of cost, and the above results are not convenient for practical application. Next, further screen the index system that plays an important role in judging drowning and build a mathematical model based on it for practical application. A 5-fold 10-fold cross-test was carried out on the previously established random forest cause of death discrimination model, as shown in Figure 5. The results showed that the prediction error of the model showed a trend of decreasing obviously and then fluctuating slightly with the increase of the number of model variables (metabolites). After cross-checking, it was found that when the model was built with 14 metabolites, the error was not statistically different from that of the model built with all metabolites. Therefore, screening according to the importance and VIP value of each metabolite is of great significance for distinguishing different causes of death. The 14 potential biomarkers are listed in Table 3. According to the box plot of 14 potential markers, it can be found that some metabolites show obvious differences in the death process of different causes of death, while other metabolites show different metabolic patterns in the early postmortem degradation process, as shown in Fig. 6. During the death process, the level of corticosterone in the drowning group increased significantly, reflecting the activation of the hypothalamus-pituitary-adrenal axis in the body, resulting in a stress response. In addition, during the drowning process, the body's cold sensory receptors dynamically respond to the sudden drop in skin temperature, accompanied by an increase in muscle tension, which increases the metabolic rate, and severe struggle in the water further increases the oxygen consumption in the event of interruption of oxygen supply, resulting in lactic acid in the drowning group. The content is higher. Most of the metabolites showed different postmortem change patterns between the two causes of death, which may be due to the different composition of microorganisms in the body during the early degradation process, and different microorganisms have different preferences for substrates.

Table 3 14 biomarkers

。

.

The 14 potential biomarkers screened contained rich information, and the random forest model was rebuilt based on the differences of these 14 metabolites among different causes of death, and its validity was tested. The results show that the accuracy of the simplified model established with 14 potential biomarkers is 95%, as shown in Table 4, which is consistent with the model established with all metabolites, AUC=0.95, as shown in Figure 7, the model is simple and effective.

Table 4 Prediction results of the grouping of validation set samples by the cause of death discriminant model established with 14 biomarkers

。

.

By performing metabolomics analysis on the heart blood samples of early water cadavers, the present invention proves the feasibility of metabolomics technology in the discrimination of drowned corpses, and further screens important biomarkers and verifies their accuracy. These results can It provides a reference for forensic practice and lays the foundation for further transformation.

Claims (1)

1. A screening method for a marker for judging whether an early underwater corpse is drowned is characterized by comprising the following specific steps:

step 1, randomly dividing heart blood samples extracted from dead rats in a drowned group and dead rats in a post-mortem and water-entering group at 0h, 6h, 12h, 18h and 24h into a training set and a verification set respectively;

step 2, performing cardioblood metabonomics detection by adopting an ultra-high performance liquid chromatography tandem mass spectrometry UHPLC-MS/MS system to obtain metabolic fingerprint spectrums of various micromolecular metabolites;

step 3, importing the training set data into R language version 3.6.1, taking sample groups as dependent variables by means of a random forest program package, taking various metabolites as independent variables, and randomly extracting training set samples in a release manner to establish a random forest classifier model, so as to finally obtain the relative importance degree of the independent variables and the prediction precision of the model;

the relative importance degree of the independent variable is expressed by taking average accuracy reduction as an index, the value is the reduction degree of the prediction accuracy of a random forest classifier model when the value of a variable is changed into a random number, the greater the importance of the variable is, the prediction accuracy of the model is expressed by the area and the accuracy of the model under the working characteristic curve of a subject on a verification set, the accuracy is the proportion of the number of correctly classified verification samples to the number of all verification samples, and the more the AUC and the accuracy are close to 1, the more reliable the model is;

step 4, cross-checking the random forest classifier model established by taking all metabolites as independent variables, establishing a simple classifier model containing less metabolites by taking the upper limit of the 95% credible interval of the cross-checking error of the classifier model established in the step 3 as reference and the same cross-checking error as standard;

step 5, selecting and determining 14 marked metabolites for judging whether the carcasses in the water are drowned according to the relative importance of the metabolites in the random forest classifier model in the step 3 and the variable projection importance value in the PLS-DA model;

the marker metabolites are: lactic acid, mevalonic acid, phenylacetylglycine, tetradecanedioic acid, inosine, malic acid, beta-hydroxybutyrate, undecanoic acid, glycylleucine, corticosterone, thiamine, acetyl-carnitine, alanylleucine and kynurenine.

Images