Biomarkers for the diagnosis of pneumoconiosis and their applications

Abstract

The present invention provides biomarkers for pneumoconiosis diagnosis and their applications, belonging to the field of biological detection technology. The biomarker is lysophosphatidylcholine (24:1), which is used in the preparation of a detection reagent for diagnosing pneumoconiosis. By combining the biomarker lysophosphatidylcholine (24:1) with lysophosphatidylinositol (16:0), ceramide (d18:1 / 23:0), sterols, and quercetin 3-(2'-galactoside), the risk of pneumoconiosis can be determined, allowing for early prevention and mitigation of the disease.

Description

Technical Field

[0001] The present invention belongs to the technical field of biological detection, and in particular relates to a biomarker for diagnosing pneumoconiosis and application thereof. Background Art

[0002] Pneumoconiosis, scientifically known as pneumoconiosis, is a typical and serious occupational disease. Due to occupational reasons, workers often inhale industrial dust, which can easily lead to dust stagnation in the lungs and cause diffuse fibrosis of lung tissue. The higher the silica content in the dust, the sooner the onset of pneumoconiosis and the more severe the disease.

[0003] According to statistics, the number of pneumoconiosis patients reached 6 million by the end of 2017. However, with current diagnostic and testing technologies, pneumoconiosis can generally only be diagnosed through lung CT scans or tissue biopsies.

[0004] The onset of pneumoconiosis varies greatly depending on individual constitution and protective measures, ranging from a few months to several years to over a decade. Due to the unique nature of pneumoconiosis's pathology and the uncertainty surrounding its onset, currently used, expensive and inefficient testing methods are becoming increasingly unsuitable for pneumoconiosis detection and diagnosis. Many patients are diagnosed with pneumoconiosis in the late stages of the disease, making subsequent treatment difficult and plunging them into prolonged pain and suffering.

[0005] Metabolomics, an emerging omics technology, is playing an increasingly important role in biological research because it can reveal the unique chemical fingerprints of cellular metabolism. As an unbiased approach to studying small molecule metabolites, metabolomics can analyze changes in endogenous metabolites to reflect the state of the organism and identify specific biomarkers or groups of markers. Finding an easily detectable biomarker and method for predicting and diagnosing pneumoconiosis is a pressing technical challenge. Summary of the Invention

[0006] In order to solve the above technical problems, the present invention provides the use of the biomarker lysophosphatidylcholine (24:1) in the preparation of a detection reagent for diagnosing pneumoconiosis.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] Application of the biomarker lysophosphatidylcholine (24:1) in the preparation of detection reagents for diagnosing pneumoconiosis.

[0009] Application of the biomarker lysophosphatidylcholine (24:1) in combination with lysophosphatidylinositol (16:0), ceramide (d18:1 / 23:0), sterols, and quercetin 3-(2'-galactoside) in the preparation of a detection reagent for diagnosing pneumoconiosis.

[0010] In the application described above, preferably, lysophosphatidylcholine (24:1) is combined with lysophosphatidylinositol (16:0), ceramide (d18:1 / 23:0), sterol Pubescenol, and quercetin 3-(2'-galactoside) to determine whether there is a risk of pneumoconiosis.

[0011] In the application described above, preferably, the content of lysophosphatidylinositol (16:0) is recorded as R36, the content of lysophosphatidylcholine (24:1) is recorded as R50, the content of quercetin 3-(2'-galactoside) is recorded as R30, and the content of sterol is recorded as R34. If any of the following conditions is met, it is determined to be pneumoconiosis. In addition to these four conditions, the rest of the conditions are not determined to be pneumoconiosis;

[0012] (1) R36<0.72, R50<0.35;

[0013] (2) R36 < 0.72, R50 ≥ 0.35, and R34 < 0.21;

[0014] (3) 0.15<R36<0.72, R50≥0.35, R34≥0.21 and R30≥0.3;

[0015] (4) R36≥0.72, R34<0.2.

[0016] In the application described above, preferably, by detecting the levels of lysophosphatidylcholine (24:1) combined with lysophosphatidylinositol (16:0), ceramide (d18:1 / 23:0), sterols, and quercetin 3-(2'-galactoside) in serum, pneumoconiosis is predicted based on the TC value in the machine learning conditional probability decision tree model composed of the above five markers: if TC ≥ 0.300, it is determined to be pneumoconiosis; if TC < 0.300, it is determined to be no pneumoconiosis.

[0017] As described above, preferably, the construction of the machine learning conditional probability decision tree model is to use the rpart package of the R language to establish the decision tree model, set the variable of whether or not pneumoconiosis is a factor variable, and the "method" parameter in the modeling is set to "class", indicating that a classification decision tree model is established; the "model" parameter is set to "False", indicating that no copy of the model framework is retained in the resampling results; the "parms" parameter is set to 1, indicating that the influence coefficient of the coefficient of variation of the prior distribution on the splitting rate is set to 1.

[0018] The beneficial effects of the present invention are:

[0019] The present invention provides a new molecular marker lysophosphatidylcholine (24:1)PC (24:1(15Z) / 0:0) and a model for distinguishing pneumoconiosis, which can be used for early detection, diagnosis and prediction of pneumoconiosis and is applied in preparing a detection kit for detecting pneumoconiosis.

[0020] The biomarkers for diagnosing cerebral pneumoconiosis provided by the present invention include the contents of lysophosphatidylcholine (24:1) combined with lysophosphatidylinositol (16:0), ceramide (d18:1 / 23:0), sterols, and quercetin 3-(2'-galactoside). The TC value in the machine learning conditional probability decision tree model composed of the five markers is used to predict pneumoconiosis, which helps to diagnose whether there is a tendency to pneumoconiosis and can be used for early prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is an S-plot comparing the normal environment control group and the susceptible pneumoconiosis environment control group under positive ion mode;

[0022] Figure 2 This is an S-plot comparing the normal environment control group and the susceptible pneumoconiosis environment control group under negative ion mode;

[0023] Figure 3 This is the S-plot comparing the pneumoconiosis susceptible environment control group and the pneumoconiosis experimental group under positive ion mode;

[0024] Figure 4 This is the S-plot comparing the pneumoconiosis susceptible environment control group and the pneumoconiosis suffering experimental group under negative ion mode;

[0025] Figure 5 are compounds with VIP>1 compared with the normal environment control group and the susceptible pneumoconiosis environment control group in positive ion mode;

[0026] Figure 6 are compounds with VIP>1 compared with the normal environment control group and the susceptible pneumoconiosis environment control group under negative ion mode;

[0027] Figure 7 The (O)PLS-DA score diagram comparing the normal environment control group and the susceptible pneumoconiosis environment control group under positive ion mode;

[0028] Figure 8 The (O)PLS-DA score diagram comparing the normal environment control group and the susceptible pneumoconiosis environment control group under negative ion mode;

[0029] Figure 9are compounds with VIP>1 in the comparison between the pneumoconiosis susceptible environment control group and the pneumoconiosis experimental group under positive ion mode;

[0030] Figure 10 are compounds with VIP>1 in the comparison between the pneumoconiosis susceptible environment control group and the pneumoconiosis experimental group under negative ion mode;

[0031] Figure 11 This is the (O)PLS-DA score diagram comparing the pneumoconiosis susceptible control group and the pneumoconiosis experimental group under positive ion mode;

[0032] Figure 12 This is the (O)PLS-DA score diagram comparing the pneumoconiosis susceptible control group and the pneumoconiosis experimental group under negative ion mode;

[0033] Figure 13 The Venn diagrams of the up- and down-regulation of preliminary markers were compared between the normal environment control group and the susceptible pneumoconiosis environment control group, and between the susceptible pneumoconiosis environment control group and the pneumoconiosis experimental group;

[0034] Figure 14 Decision tree probability model diagram for the final markers obtained by comparing the susceptible pneumoconiosis control group with the pneumoconiosis experimental group (variables are R36+R20+R34+R30+R50);

[0035] Figure 15 Evaluation diagram of the decision tree probability model established for the final markers obtained by comparing the susceptible pneumoconiosis control group with the pneumoconiosis experimental group (variables are R36+R20+R34+R30+R50). DETAILED DESCRIPTION

[0036] The following examples are used to further illustrate the present invention, but should not be construed as limiting the present invention. Without departing from the spirit and substance of the present invention, modifications or substitutions made to the present invention all fall within the scope of the present invention.

[0037] Unless otherwise specified, the technical means used in the examples are conventional means well known to those skilled in the art.

[0038] Example 1

[0039] A total of 196 samples were used to establish the model, with an age range of 30 years and above, including 50 controls without pneumoconiosis who lived or worked in a normal environment, 85 controls without pneumoconiosis who lived or worked in a pneumoconiosis-susceptible environment, and 61 patients with pneumoconiosis.

[0040] Among the control group without pneumoconiosis who lived or worked in a normal environment and the control group without pneumoconiosis who lived or worked in an environment susceptible to pneumoconiosis, lung function indicators were normal and chest X-ray imaging showed no abnormalities.

[0041] Among patients with pneumoconiosis, lung function indicators are abnormal, and chest X-ray imaging tests show abnormalities.

[0042] Sample collection: The test was conducted on sera from people who were clinically evaluated as not suffering from pneumoconiosis living or working in a normal environment, people who were living or working in a pneumoconiosis-susceptible environment and patients with pneumoconiosis.

[0043] Sample preparation

[0044] Serum samples collected from the above-mentioned populations were thawed on ice. 200 μL of serum was extracted with 600 μL of pre-chilled isopropanol, vortexed for 1 minute, and incubated at room temperature for 10 minutes. The extraction mixture was then stored at −20°C overnight and centrifuged at 12,000 rpm in a low-temperature refrigerated centrifuge (E3116R, ESSENSCIEN, USA) for 20 minutes. 260 μL of the supernatant was transferred to a new centrifuge tube and added with 130 μL of isopropanol, 130 μL of acetonitrile, and 65 μL of ultrapure water, adjusting the sample isopropanol / acetonitrile / water ratio to 2.5:1:1 (v:v). Samples were stored at −80°C prior to LC-MS analysis. In addition, 10 μL of each extraction mixture was combined to prepare a pooled QC sample.

[0045] The reagents used in the present invention: isopropanol, formic acid, acetonitrile, ammonium formate, LC-MS grade manufacturers are all Fisher.

[0046] 2. Ultra-high performance liquid chromatography-mass spectrometry for lipidomics

[0047] The samples were analyzed by ultra-performance liquid chromatography (UPLC; model: ACQUITY UPLC I-Class system; manufacturer: Waters, Manchester, UK) connected to a Xevo-G2XS high-resolution time-of-flight (QTOF) mass spectrometer (Waters) with ESI. A CQUITY UPLC BEH C18 column (2.1 × 100 mm, 1.7 μm, Waters) was used. Mobile phase A consisted of 10 mM ammonium formate-0.1% formic acid-acetonitrile-ultrapure water (prepared by weighing 0.63 g of ammonium formate and 10 g of formic acid, dissolving them in acetonitrile-water solution (acetonitrile:water, 60:40, v / v), and calibrating the volume to 1000 mL). Mobile phase B consisted of 10 mM ammonium formate-0.1% formic acid-isopropanol-acetonitrile (prepared by weighing 0.63 g of ammonium formate and 10 g of formic acid, dissolving them in isopropanol-acetonitrile solution (isopropanol:acetonitrile, 90:10, v / v), and calibrating the volume to 1000 mL). Prior to large-scale studies, pilot experiments with elution periods of 10, 15, and 20 minutes were performed to assess the potential effects of mobile phase composition and flow rate on lipid retention. In the PIM, abundant lipid precursor ions and fragments resolved in the same order, with similar peak shapes and ion intensities. Furthermore, a pooled QC sample with a 10-minute elution period also exhibited similar precursor and fragment base peak intensities as the test samples. The mobile phase flow rate was 0.4 mL / min. The column was initially eluted with 40% B, followed by a linear gradient to 43% B over 2 minutes, and then the percentage of B was increased to 50% over 0.1 minute. Over the next 3.9 minutes, the gradient was further increased to 54% B, and the amount of B was then increased to 70% over 0.1 minute. In the final portion of the gradient, the amount of B increased to 99% over 1.9 minutes. Finally, the solution B was returned to 40% over 0.1 minute, and the column was equilibrated for 1.9 minutes before the next injection. Each injection volume was 5 μL, and lipids were detected in both positive and negative modes using a Xevo-G2XS QTOF mass spectrometer, acquiring from m / z 50 to 1200 with an acquisition time of 0.2 seconds per injection. The ion source temperature was 120°C, the desolvation temperature was 600°C, and the gas flow rate was 1000 L / h, with nitrogen as the flowing gas. The capillary voltage was 2.0 kV (+) and the cone voltage was 1.5 kV (-), with a cone voltage of 30 V. Leucine enkephalin was used for standard mass determination, and calibration was performed using sodium formate solution. Samples were randomly ordered. A QC sample was injected and analyzed every 10 samples to investigate data reproducibility.

[0048] Data acquisition software (MassLynx4.1; manufacturer: Waters) was used for data acquisition and result analysis:

[0049] 3. Using traditional statistics to find serum differential substances

[0050] Progenesis QI was used to convert mass spectrometry data into a statistically applicable data form. Orthogonal partial least squares discriminant analysis (OPLS-DA) combined the orthogonal signal correction (OSC) and PLS-DA (partial least squares discriminant analysis) methods to screen the difference variables by removing irrelevant differences. Figure 1 、 Figure 2 This is an S-plot comparing the normal environment control group and the susceptible pneumoconiosis environment control group under positive and negative ion modes (A represents positive ions, B represents negative ions). Figure 3 、 Figure 4 This is an S-plot comparing the pneumoconiosis-susceptible control group and the pneumoconiosis-affected experimental group under positive and negative ion modes (A represents positive ions, B represents negative ions). The horizontal axis represents the cocorrelation coefficient between the principal component and the metabolite, and the vertical axis represents the correlation coefficient between the principal component and the metabolite. When p < 0.05 was met, 2099 substances were found in the positive ion mode and 1383 substances were found in the negative ion mode when the normal control group was compared with the pneumoconiosis-susceptible control group. When p < 0.05 was met, 2109 substances were found in the positive ion mode and 1383 substances were found in the negative ion mode when the normal control group was compared with the pneumoconiosis-susceptible control group.

[0051] 4. Use multivariate statistics to find significantly different substances in serum

[0052] Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) combines the Orthogonal Signal Correction (OSC) and PLS-DA (Partial Least Squares Discriminant Analysis) methods to screen out differential variables by removing irrelevant differences. Figure 5 and Figure 6 As shown: VIP value is the variable importance projection of the first principal component of PLSDA when comparing the normal environmental control group with the susceptible pneumoconiosis environmental control group under positive and negative ion modes. VIP>1 is usually used as a common evaluation standard in metabolomics and as one of the criteria for screening differential metabolites. Figure 7 and Figure 8 The following is a score diagram of the first and second principal components of the two groups under the positive and negative ion modes, the normal environmental control group (CK) and the susceptible pneumoconiosis environmental control group (CFD), obtained by dimensionality reduction. The horizontal axis represents the difference between the groups, and the vertical axis represents the difference within the groups. The results of the two groups are well separated, indicating that this scheme can be used. The modeling of the susceptible pneumoconiosis environmental control group (CFD) and the pneumoconiosis experimental group (CFB) is similar (VIP Figure: Figure 9 and Figure 10 , principal component plot: Figure 11 and Figure 12Under the conditions of p<0.05 and VIP>1, there were 92 differences in the positive ion mode and 159 differences in the negative ion mode between the normal environment control group (CK) and the susceptible pneumoconiosis environment control group (CFD), for a total of 261 substances. There were 45 differences in the positive ion mode and 27 differences in the negative ion mode between the susceptible pneumoconiosis environment control group (CFD) and the pneumoconiosis experimental group (CFB), for a total of 72 substances. Figure 5 and Figure 6 、 Figure 9 and Figure 10 This indicates that when the three groups were compared pairwise, there were many significantly enriched compounds (VIP value greater than 1) compared between the CK group and the CFD group, and between the CFD and CFB groups, which may be due to interference factors. Figure 7 and Figure 8 、 Figure 11 and Figure 12 This shows that when the three groups are compared pairwise, the comparison between the CK group and the CFD group, and between the CFD and CFB groups can achieve good modeling effects, and the sample information between the two groups can be effectively extracted.

[0053] 5. Elimination of confounding factors of different substances

[0054] Further comparison was made for the up-regulated or down-regulated compounds. Due to the fact that pneumoconiosis cannot be acquired in a non-susceptible environment, all compounds with differences between the CFD group and the CFB group were eliminated as confounding factors. Figure 13 In the comparison between the CFD group and the CFB group, the overlapping parts of the differences between the CK group and the CFD group were removed, and finally 26 highly specific up-regulated differences in the CFB group and 27 highly specific down-regulated differences in the CFB group were obtained.

[0055] 6. Feature Importance Filter Analysis

[0056] To further narrow the scope, 53 compounds were subjected to a filtering score analysis to reflect the degree to which a single indicator provides information about the outcome, with higher scores indicating greater importance. With a reproducible seed value of I = 2193 and a filtering score threshold of 3.5, five compounds were ultimately identified. The results are shown in Table 1.

[0057] Table 1 Filter analysis of pneumoconiosis-related lipids

[0058]

[0059] 7. Ten-fold cross-validation results of internal population

[0060] In order to improve the biodiagnostic effect of variable compounds, it is necessary to find a suitable model based on the above-mentioned biomarkers to carry out the next step of analysis. Due to the existence of nonlinearity in the real world, a semi-parametric decision tree probability model is selected in the embodiment, and the model consists of nodes and directed edges. There are two types of nodes: internal nodes and leaf nodes, where internal nodes represent a feature or attribute, and leaf nodes represent a class. Generally, a decision tree contains a root node, several internal nodes and several leaf nodes. The leaf node corresponds to the decision result, and each other node corresponds to an attribute test. The sample set contained in each node is divided into child nodes according to the result of the attribute test. The root node contains the full set of samples, and the path from the root node to each leaf node corresponds to a judgment test sequence.

[0061] The entire population was randomly divided into 10 groups, one of which was selected as the validation set, and the others as the training set. This process was repeated ten times to identify the optimal variable combination. The internal stability of the five compound combinations during the decision tree model construction process was examined, including the average values of AUC, sensitivity, and specificity, and statistical significance was calculated. The results are shown in Table 2 below, where the sequence number corresponds to the number of internal validations. The independent variables for the model were the five compounds mentioned above, and the dependent variable was the outcome variable, TC.

[0062] Table 2

[0063]

[0064] The decision tree model formed by the combination of five compounds has excellent internal stability, and the average AUC is stable at around 0.846.

[0065] Similarly, the internal stability of the four compound combinations except R50 during the decision tree model construction process was examined, including taking the average value of AUC, sensitivity, and specificity, and performing statistical significance calculations. The results are shown in Table 3 below:

[0066] Table 3

[0067]

[0068] The decision tree model formed by the four-compound combination showed poor internal stability, with the average AUC remaining stable at around 0.684. In other words, the presence of compound R50 made it possible to establish an internally stable model, and compound R50 also played a crucial role in the marker combination.

[0069] Based on the above analysis, a decision tree model was established using the original five compounds. The model diagram is shown in the figure below. Figure 14Among them, the content of lysophosphatidylinositol (16:0) is recorded as R36, the content of lysophosphatidylcholine (24:1) is recorded as R50, the content of quercetin 3-(2'-galactoside) is recorded as R30, and the content of sterol is recorded as R34. Taken together, the following four situations will be judged as pneumoconiosis patients, as follows: when the R36 value is less than 0.72, the R50 value is less than 0.35; when the R36 value is less than 0.72, the R50 value is greater than or equal to 0.35 and the R34 value is less than 0.21; when the R36 value is less than 0.72, the R50 value is greater than or equal to 0.35 and the R34 value is greater than or equal to 0.21 and the R30 value is greater than or equal to 0.3 and the R36 value is less than 0.15; when the R36 value is greater than or equal to 0.72 and the R34 value is less than 0.2. In all other cases except these four, the test subjects are determined not to be pneumoconiosis patients.

[0070] That is to say, the content of each substance measured is recorded as R36 for lysophosphatidylinositol (16:0), R50 for lysophosphatidylcholine (24:1), R30 for quercetin 3-(2'-galactoside), and R34 for sterol. If any of the following conditions are met, the subject is judged to be a pneumoconiosis patient. In all other cases except these four, the subject is not judged to be a pneumoconiosis patient.

[0071] (1) R36<0.72, R50<0.35;

[0072] (2) R36 < 0.72, R50 ≥ 0.35, and R34 < 0.21;

[0073] (3) 0.15<R36<0.72, R50≥0.35, R34≥0.21 and R30≥0.3;

[0074] (4) R36≥0.72, R34<0.2.

[0075] 8. External Dataset, Decision Tree Model Verification

[0076] The accuracy of the above results was verified by using a data set from an external population, and the corresponding ROC curve was plotted. The results are as follows:

[0077] Validation population: 204 individuals (external population), sampled using the same criteria as the sample population described above. 101 controls without pneumoconiosis who live or work in a pneumoconiosis-susceptible environment, and 103 individuals with pneumoconiosis. Decision tree model validation was performed.

[0078] The variables in the model are the above five metabolites R36+R20+R34+R30+R50. A decision tree model was established based on the compound content values. The decision tree model was established using the rpart package of R language. The variable of whether or not pneumoconiosis was present was set as a factor variable. The "method" parameter in the modeling was set to "class", indicating that a classification decision tree model was established; the "model" parameter was set to "False", indicating that no copy of the model framework was retained in the resampling results; the "parms" parameter was set to 1, indicating that the coefficient of variation of the prior distribution on the division rate was set to 1; The model evaluation diagram is shown in the figure below. Figure 15 .

[0079] Sensitivity = 0.820

[0080] Specificity = 0.812

[0081] Accuracy = 0.815

[0082] Thresholds = 0.300

[0083] Pneumoconiosis is predicted by introducing the TC value into the machine learning conditional probability decision tree model based on the five marker detection data of the sample. If TC ≥ 0.300, the sample is judged to have pneumoconiosis; if TC < 0.300, the sample does not have pneumoconiosis.

[0084] The data showed that lysophosphatidylcholine (24:1), combined with lysophosphatidylinositol (16:0), ceramide (d18:1 / 23:0), sterols, and quercetin 3-(2'-galactoside), exhibited very high diagnostic capabilities and could be used in clinical test kits in the future.

[0085] Comparative analysis of sample information revealed that, compared with the pneumoconiosis-susceptible environmental control group, the above five biomarkers, including lysophosphatidylinositol (16:0), ceramide (d18:1 / 23:0), sterols, and lysophosphatidylcholine (24:1), showed a downward trend in the pneumoconiosis patient group, while the opposite trend was observed for quercetin 3-(2'-galactoside).

Claims

1. Use of the biomarker lysophosphatidylcholine (24:1) in combination with lysophosphatidylinositol (16:0), ceramide (d18:1 / 23:0), sterols and / or quercetin 3-(2'-galactoside) in the preparation of a detection reagent for diagnosing pneumoconiosis, characterized in that: By detecting the contents of lysophosphatidylcholine (24:1) combined with lysophosphatidylinositol (16:0), ceramide (d18:1 / 23:0), sterols, and quercetin 3-(2'-galactoside) in serum, the content of lysophosphatidylinositol (16:0) is recorded as R36, the content of lysophosphatidylcholine (24:1) is recorded as R50, the content of quercetin 3-(2'-galactoside) is recorded as R30, and the content of sterols is recorded as R34. If any of the following conditions is met, it is determined to be pneumoconiosis. In addition to these four conditions, the rest of the conditions are not determined to be pneumoconiosis; (1) R36<0.72, R50<0.35; (2) R36 < 0.72, R50 ≥ 0.35, and R34 < 0.21; (3) 0.15<R36<0.72, R50≥0.35, R34≥0.21 and R30≥0.3; (4) R36≥0.72, R34<0.2; Alternatively, pneumoconiosis can be predicted based on the TC value in the machine learning conditional probability decision tree model composed of the above five markers: if TC ≥ 0.300, the patient is diagnosed with pneumoconiosis; if TC < 0.300, the patient is not diagnosed with pneumoconiosis; The machine learning conditional probability decision tree model was constructed using the rpart package of the R language. The variable of whether or not pneumoconiosis was present was set as a factor variable. The "method" parameter in the modeling was set to "class", indicating that a classification decision tree model was established; the "model" parameter was set to "False", indicating that no copy of the model framework was retained in the resampling results; and the "parms" parameter was set to 1, indicating that the coefficient of variation of the prior distribution on the splitting rate was set to 1.

Images