CN113138250B

CN113138250B - Non-diagnostic method for typing covid-19 grade by using characteristic urine protein and application

Info

Publication number: CN113138250B
Application number: CN202110441023.6A
Authority: CN
Inventors: 郭天南; 丁璇; 刘威; 朱怡
Original assignee: Westlake University
Current assignee: Westlake University
Priority date: 2021-04-23
Filing date: 2021-04-23
Publication date: 2021-12-17
Anticipated expiration: 2041-04-23
Also published as: CN113138250A

Abstract

The invention provides a non-diagnostic method for typing covid-19 light and heavy grades by using characteristic urine proteins and application thereof, the applicant defines 20 characteristic urine proteins through proteomic analysis, and can obtain a predicted value by inputting the relative expression amount into typing on the basis of obtaining the relative expression amount of the 20 characteristic urine proteins in urine of a subject, and the predicted value is associated with the covid-19 typing result of the subject.

Description

Non-diagnostic method for typing covid-19 grade by using characteristic urine protein and application

Technical Field

The invention relates to the field of biological medicine, in particular to a non-diagnostic method for typing covid-19 grade by using characteristic urine protein and application thereof.

Background

The clinical symptoms of Covid-19 mainly include fever, dry cough, hypodynamia and the like, a few patients have upper respiratory tract and digestive tract symptoms such as nasal obstruction, watery nasal discharge, diarrhea and the like, the abuse of Covid-19 threatens the health of hundreds of millions of people in the world, and according to the report of the world health organization, more than 8300 million people are infected with SARS-CoV-2 all over the world by 1 month and 4 days of 2021 year, and the number of dead people is more than 180 million. In clinical practice, about 80% of Covid-19 patients are non-severe cases, the symptoms are mild, the prognosis is good, the rest 20% of Covid-19 patients are severe and need special care including oxygen inhalation, artificial ventilation and other operations, however, the disease development speed of severe Covid-19 patients is extremely rapid, dyspnea appears after 1 week in most severe cases of Covid-19, severe patients can rapidly progress to acute respiratory distress syndrome, septic shock, metabolic acidosis which is difficult to correct, coagulation dysfunction, multi-organ failure and other conditions, and therefore, the early discovery of severe Covid-19 patients has very wide significance.

At present, the detection of Covid-19 mainly depends on throat swab nucleic acid detection, and the detection method can only detect the existence of virus, but cannot indicate the severity of the disease of a patient, and further cannot predict whether the patient possibly has tissue damage. That is, the current lightweight typing of Covid-19 can only be judged by the clinical symptoms of the patient, and most likely, the manual intervention optimal for the clinical symptoms is missed.

Disclosure of Invention

The invention aims to provide a non-diagnostic method for typing covid-19 grade by using characteristic urine protein and application thereof, in particular, the relative expression quantity of the characteristic urine protein in urine of a subject is detected, the relative expression quantity is input into a typing model to obtain a predicted value of a covid-19 typing result, the predicted value can be used for typing the covid-19 grade and monitoring the damage degree of kidney tissues of the subject suffering from the covid-19, and the characteristic urine protein is as follows: VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84, and BTNL 3.

The characteristic urine protein provided by the scheme is used in combination for grading covid-19 weight: grading the weight of covid-19 based on the predicted value; for the prevention and treatment of kidney tissue damage caused by covid-19: if the classification is heavy covid-19, the patient is judged to have the verified kidney tissue damage in advance; for monitoring the worsening or improvement of a condition in a subject: monitoring the changing status of covid-19 disorder based on the change in the predicted value; the application of the covid-19 in preparing the targeted medicine for treating the heaviness and the heaviness of covid-19 provides support for: whether the medicine has the treatment effect is verified based on the predicted value, and the characteristic urine protein of the scheme has the greatest difference in light and heavy typing, and can be used as the research focus of future medicine targets. Is beneficial to more targeted treatment and medication in the clinical diagnosis and treatment process, and has great significance for improving the prognosis of patients.

In addition, the method is used for prejudging the covid-19 typing result by detecting the relative expression quantity of the characteristic urine protein in the urine of a subject, compared with a blood sampling or throat swab nucleic acid detection mode, the urine sampling has the characteristics of convenience in sampling and high patient compliance, and the detection can be completed only by using a trace urine sample (500 ul).

In a first aspect, the present invention provides a non-diagnostic method for characterizing the classification of covid-19 grade by urinary protein, comprising the steps of: detecting the relative expression amount of characteristic urine proteins in urine of a subject, inputting the relative expression amount of the urine proteins into a typing model to obtain a predicted value, and if the predicted value is lower than a set threshold value, typing the subject to be severe, wherein the characteristic urine proteins comprise VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84 and BTNL 3.

Only 500. mu.L of urine of the subject needs to be taken, and specific methods for detecting the relative expression amount of the characteristic urine protein in the urine of the subject include, but are not limited to: and the detection is carried out by utilizing a tmt labeled urine proteomics analysis method, and other target or non-target proteomics analysis methods.

Wherein the urine proteomics analysis method using the tmt marker comprises:

inactivating and sterilizing the urine sample, adding acetone to precipitate protein overnight, adding urea to denature after resuspension, reducing and alkylating the protein lysate, diluting with TEAB after incubation for a set time, and stopping reaction with trypsin and mixed TEA; washing and desalting the peptide fragments subjected to pancreatin digestion by using a desalting column, then marking the peptides by using tmt, dissolving the peptides, analyzing by using LC-MS/MS to obtain mass spectrum data, and acquiring quantitative data of the peptides and the proteins based on the mass spectrum data.

Specifically, in the present scheme, the specific means are as follows: urine samples were inactivated and sterilized at 56 ℃ for 30 min, then precipitated overnight with cold acetone (urine: acetone ═ 1:4, v/v, -20 ℃), resuspended with 100. mu.L TEAB, denatured by addition of 100. mu.L 10M urea, and protein lysates reduced and alkylated with 10mm tris (2-carboxyethyl) phosphine (TCEP) and 40mm Iodoacetamide (IAA) and incubated at 30 ℃ for 30 min in the absence of light. After further dilution with 200. mu.L of 100mM TEAB, the mixture was digested with 5. mu.g trypsin and 1. mu.g Lys-C for 12 hours at 32 ℃. The reaction was stopped by adding 30. mu.L of 10% trifluoroacetic acid (TFA). These tryptic peptides were washed and desalted using a desalting column and the peptides were labeled with TMTpro 16plex (thermo Fisher) according to the manufacturer's instructions. The urine polypeptide was again dissolved with 2% aqueous acetonitrile (V/V) and then analyzed by LC-MS/MS to obtain mass spectra data. And identifying and quantifying the obtained mass spectrum data by using commercial Proteome discover 2.4.1.15 software, and outputting the identification and quantification results of the polypeptide and the protein.

The typing model is obtained by training a machine learning model by taking the relative expression amount of characteristic urine proteins of different weight-weight covid-19 subjects as training samples, and the predicted value of the typing model is used for characterizing the weight of the covid-19 of the subjects, wherein the characteristic urine proteins comprise VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84 and BTNL 3.

Specifically, in the scheme, the machine learning model is constructed by a random forest algorithm (4.6.14 version, R software package), the relative expression quantities of the 20 characteristic urine proteins in 39 light patients and 11 heavy patients are used as the raw data of the input model, and the main parameters of the model construction are set as: ntree is 1000, nodesize is 1, others use default parameters.

In order to improve the prediction precision of the typing model, the training samples are at least selected from the following: the relative expression level of urine protein of a patient with severe covid-19 and the relative expression level of urine protein of a patient with mild covid-19, and respectively marking the type of the sample as light or heavy; the trained typing model can output a predicted value based on the input relative expression quantity of the urine protein, and the higher the predicted value is, the more the typing of the covid-19 is favored to be light, otherwise, the more the typing is favored to be heavy.

In this scenario, the threshold for the prediction value is 0.5, i.e., if the prediction value is less than 0.5, the subject is considered to be covid-19 severe.

In a second aspect, the invention provides a use of characteristic urine protein as a detection target in preparing a kit for typing the weight of covid-19 of a subject, wherein the kit contains a reagent for detecting the relative expression amount of the characteristic urine protein; detecting the relative expression quantity of the characteristic urine protein in the urine of the subject by using the kit, wherein the relative expression quantity of the characteristic urine protein is related to the covid-19 typing result of the subject, and the characteristic urine protein is as follows: VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84, and BTNL 3.

Wherein the relative expression of characteristic urine proteins correlates with subject covid-19 typing results: and inputting the relative expression quantity of the characteristic urine protein into the typing model, and if the predicted value output of the typing model is lower than a set threshold value, predicting that the test subject is covid-19 as heavy.

In a third aspect, the present invention provides an application of characteristic urine protein in preparing a reagent for typing covid-19 weight of a subject, detecting a relative expression amount of the characteristic urine protein in urine of the subject, inputting the relative expression amount into a typing model to predict a covid-19 typing result of the subject, wherein the characteristic urine protein is: VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84, and BTNL 3.

In a fourth aspect, the scheme provides a typing model for typing the weight and weight of covid-19 of a testee, wherein the relative expression quantity of the characteristic urine protein of urine of different weight and weight covid-19 testees is used as a training sample to train a machine learning model to obtain, and the characteristic urine protein is as follows: VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84 and BTNL3, and the predictive value of the typing model characterizes covid-19 lightness and heaviness of the subject.

Compared with the prior art, the scheme has the characteristics that: 20 characteristic urine proteins which can be used for indicating the weight and the weight of the Covid-19 are selected through analysis of proteomic data, a predicted value is obtained through prediction by inputting relative expression quantities of the 20 characteristic urine proteins into a typing model, and the weight of the Covid-19 can be typed based on the predicted value, so that typing of a patient can be judged as early as possible, more targeted treatment and medication in a clinical diagnosis and treatment process are facilitated, great significance is brought to improvement of prognosis of the patient, and the characteristic urine proteins can be used as potential treatment targets for follow-up research.

Drawings

FIG. 1 is a graph showing the production of polypeptides in urine from critically ill, non-critically ill COVID-19 patients, non-COVID-19 patients and healthy donors.

FIG. 2 is a graph of 20 characteristic urine protein data obtained by screening.

FIG. 3 is a graph showing the results of data output from the typing model after the relative expression level of urine protein in typing experiment 1 was input.

FIG. 4 shows the relative expression levels of urine protein in 13 covid-19 patients classified in experiment 1.

FIG. 5 is a graph of data for 20 characteristic urine proteins required for control experiment 2.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments that can be derived by one of ordinary skill in the art from the embodiments given herein are intended to be within the scope of the present invention.

The inventor carries out omics analysis on the protein of covid-19 urine to show that 20 characteristic urine proteins are selected as markers:

analyzing a sample: urine samples from 50 covid-19 patients (39 non-severe and 11 severe), 23 healthy donors were obtained from Taizhou Hospital, Zhejiang.

The analysis process comprises the following steps:

sample preparation: an aliquot of the patient urine was taken, and the urine specimen was inactivated and sterilized at 56 ℃ for 30 minutes, then precipitated overnight with cold acetone (urine: acetone ═ 1:4, v/v, -20 ℃), resuspended with 100. mu.L TEAB, denatured by adding 100. mu.L of 10M urea, and the protein lysate was reduced and alkylated with 10mm tris (2-carboxyethyl) phosphine (TCEP) and 40mm Iodoacetamide (IAA) and incubated at 30 ℃ for 30 minutes in the dark. After further dilution with 100mM TEAB 200. mu.L, digestion was carried out with 2. mu.g trypsin at 32 ℃ for 4 hours, followed by another 2. mu.g trypsin at 32 ℃ for 12 hours. The reaction was stopped by adding 30. mu.L of 10% trifluoroacetic acid (TFA), and the tryptic peptide fragments were washed with a desalting column to remove salt, and then labeled with TMTpro 16plex according to the manufacturer's instructions.

Nanoliter liquid phase-high resolution mass spectrometry:

urine polypeptides re-dissolved with 2% aqueous acetonitrile (V/V) were analyzed by LC-MS/MS coupling Q-active HF-X hybridization quadrupole orbitrap in Data Dependent Acquisition (DDA) mode using the same LC system. The experiments in this scheme were run in batch design to minimize the impact of batch effects on proteomics data. The applicant randomly divides four groups of different samples into 6 batches for TMTpro 16plex marking, the number of the samples in each batch is the same, and the four groups of different samples refer to: urine samples from COVID-19 critically ill patients, COVID-19 non-critically ill patients, non-COVID-19 patients, and healthy donors; for each batch of TMT samples, a naflow DIONEX UltMate 3000RSLCnano system (Thermo Fisher Scientific, San Jose, USA) and an Xbridge Peptide BEH C18 chromatography column(s) was used

5 μm.times.4.6 mm.times.250 mm) (Waters, Milford, MA, USA). Samples were separated with a gradient of 5% to 35% Acetonitrile (ACN) in 10mM ammonia (pH 10.0) at a flow rate of 1 mL/min. The TMT-tagged peptides were separated by the system into 60 fractions, which were further combined into 30 fractions. After spin-drying, the 30 fractions were redissolved with 2% ACN/0.1% Formic Acid (FA), and the redissolved polypeptides were analyzed by LC-MS/MS. Each fraction was analyzed on-line using a Data Dependent Acquisition (DDA) mode using a naflow DIONEX UltiMate 3000RSLCnano system (Thermo Fisher Scientific, San Jose, USA) in combination with QE-HFX high resolution mass spectrometry (Thermo Fisher Scientific, San Jose, USA), with the sample first loaded onto a pre-loaded column (3 μm,

20mm 75 μm i.d.) and then the sample loaded on the pre-loaded column was washed into an analytical column (1.9 μm,120a,150 mm 75 μm certificate) at a flow rate of 300 nL/min for further on-line separation, with an analytical time of 35 min and an LC gradient of from 5% to 28%Wash B (buffer A was 2% ACN, 98% H2O (containing 0.1% FA), and buffer B was 98% ACN (containing 0.1% FA)). All reagents were MS grade. The m/z range of MS1 in terms of mass spectrum parameters is 350-1800, the resolution is 60,000(200m/z), the AGC is 3e6, and the maximum ion implantation time (max IT) is 50 MS. The first 15 precursor ions were selected for MS/MS secondary fragmentation with a resolution of 45000(200m/z), AGC of 2e5, and max IT of 120 MS. The isolation window of the selected precursor was 0.7 m/z. The mass spectral data was analyzed using a Proteome scanner (version 2.4.1.15, Thermo Fisher Scientific) and protein database (downloaded from UniProtKB).

In this protocol, the enzyme is set to trypsin, with two deletions of cleavage tolerance. Static modifications are cysteine aminomethylation (+57.021464), lysine residues and peptide N-terminal TMTpro (+304.207145), and variable modifications are methionine oxidation (+15.994915) and peptide N-terminal acetylation (+ 42.010565). The mass deviation of the precursor ions was set to 10ppm and the mass deviation of the fragmented ions was set to 0.02 Da.

Results analysis that the peptide content in urine of the patients with severe and non-severe covid-19 is obviously higher than that of healthy volunteers can be seen as shown in figure 1, and the results show that the urine of the patients with covid-19 has increased protein amount and has potential proteinuria behavior characteristics. In fig. 1, healthy control groups refer to: a healthy donor; light patients refer to: 39 cases were not severe; heavy patients refer to: 11 cases with severe illness.

And the applicant identified 19,732 polypeptides and 3854 proteins in urine and obtained their respective relative quantitative results. The median Coefficient of Variance (CV) of the Quality Control (QC) samples for proteome data was 13%, indicating that the data quality of the samples selected for this protocol was high. Specifically, the data indicate that the data of the scheme is small in systematic error, and the selection of the 20 urine proteins cannot be mistaken due to the systematic error. The sample is specially inserted and provided with a QC sample in the identification process, and the data of the QC sample is used for proving the reliability of the analysis method of the scheme, so that the 20 screened proteins in the scheme are proved to be not derived from system errors but real differences of the samples.

On the premise of proving that the sample data selected by the scheme has high quality, the sample data of 50 covid-19 patients is continuously selected as an analysis sample, and the difference of the covid-19 mild disease and severe disease is evaluated by using a machine learning method on the proteome level, so as to finally locate 20 kinds of characteristic urine proteins.

The process of data mining is as follows:

the R software package of random forest (4.6.14 edition) is adopted, the relative expression quantity of 3854 proteins identified in urine is used as variable input in urine data, and the parameters of random forest are set as follows: ntree ═ 5000 and nodesize ═ 1, and the other parameters were set by default, and the proteins obtained by screening were sorted by Mean Decrease (average reduction precision) output with these proteins. Average reduction precision top 20 urine proteins, shown in figure 2, that are primarily involved in cell adhesion, cell development, secretion, digestion and extracellular matrix or structural tissue-related biological processes are VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84 and BTNL 3.

The following is presented for urine protein 20:

the VPS36 protein, which is a component of the endocytic sorting transport complex (ESCRT) II, plays an important role in the viral teething process.

CEL protein promotes the production of large numbers of chylomicrons in the intestine, catalyzing the hydrolysis of a variety of substrates, including cholesterol esters, phospholipids, lysophospholipids, di-and triacylglycerols, and fatty acid esters of hydroxy fatty acids (FAHFA).

The protein encoded by FREM2 localizes to the basement membrane and may itself form a ternary complex that plays a role in epidermal-dermal interactions. The protein is important for the integrity of the skin and renal epithelium.

TNR encodes a member of the extracellular matrix glycoprotein tenascin family. The protein may play a role in the regulation of neurite outgrowth, neuronal cell adhesion and sodium channel function.

PTGFRN (prostaglandin F2 receptor inhibitor) inhibits the binding of prostaglandin F2-alpha (PGF2-alpha) to its specific FP receptor by reducing the number of receptors rather than the affinity constant. Is associated with prostate function.

MLEC proteins have affinity for Glc2Man9GlcNAc2(G2M9) N-glycans and are involved in regulating glycosylation in the endoplasmic reticulum. The protein was also shown to interact with ribonucleoprotein I and may be involved in directing the degradation of misfolded proteins. The pathways involved include the innate immune system and protein metabolism.

ADGRL1 is a member of the laterophilin subfamily of G-protein coupled receptors (GPCRs) and may play a role in cell adhesion and signal transduction.

OSTN is a hormone that acts as a regulator of dendritic growth during development of the cerebral cortex.

ICOSLG is a ligand for the T cell surface receptor ICOS. ICOSLG can induce B cell proliferation as a costimulatory signal for T cell proliferation and cytokine secretion. ICOSLG plays an important role in regulating secondary immune responses by stimulating memory T cell function.

One of the three natriuretic peptide receptors NPR3 regulates blood volume and pressure, pulmonary hypertension and cardiac function, as well as certain metabolic and growth processes.

MELTF is a cell surface glycoprotein found on melanoma cells. The protein has sequence similarity and iron binding properties with transferrin superfamily members.

PRSS2 belongs to the trypsin family of serine proteases and encodes the anionic trypsinogen. It is part of the trypsinogen cluster located within the T cell receptor β locus. This protein is present in high amounts in the pancreatic juice and its upregulation is characteristic of pancreatitis.

SERPINI1 is a member of the serine protease inhibitor superfamily. This protein is secreted mainly by axons in the brain and preferentially reacts with tissue-type plasminogen activator and inhibits tissue-type plasminogen activation. It is thought to play a role in the regulation of axonal growth and in the development of synaptic plasticity.

MADCAM1 is an endothelial cell adhesion molecule that interacts with leukocyte beta7 integrin lpam1, l-selectin, and VLA-4 on bone marrow cells to direct leukocytes into the mucosa and inflamed tissues. A decrease in urine MADCAM1 levels in patients with COVID-19 may indicate that the process of leukocyte extravasation to sites of inflammation is affected.

CDH22 is a member of the cadherin superfamily. During the development and maintenance of the brain and neuroendocrine organs, it is possible that it plays an important role in morphogenesis and tissue formation in neural and non-neural cells.

CDH19 protein is a calcium-dependent cell-cell adhesion glycoprotein, loss of cadherin may be associated with the development of cancer, and diseases associated with CDH19 include charge syndrome and small eye disease.

TRHDE (thyroid-stimulating hormone-releasing hormone-degrading enzyme) is an extracellular peptidase which specifically cleaves and inactivates the neuropeptide thyroid-stimulating hormone-releasing hormone. The TRHDE-associated diseases include Fraser syndrome 1.

SPOCK1 may play a role in cell-cell and cell-matrix interactions. May contribute to various neuronal mechanisms of the central nervous system. Diseases associated with SPOCK1 include deafness, autosomal recessive inheritance 101 and gingival overgrowth.

CD84 is an autorigand receptor of the Signaling Lymphocyte Activating Molecule (SLAM) family, involved in the regulation and interconnection of innate and adaptive immune responses. . Can mediate Natural Killer (NK) cytotoxicity (by similarity) dependent on SH2D1A and SH2D1B, and can additionally activate the proliferative response of T cells. In macrophages, LPS-induced MAPK phosphorylation and NF- κ B activation are enhanced and LPS-induced cytokine secretion is modulated. Diseases associated with CD84 include lymphoproliferative syndrome and selective immunoglobulin deficiency.

An important paralogue homolog of BTNL3 is BTNL8, which may stimulate a primary immune response. T cells stimulated by action on the TCR/CD3 complex stimulate their proliferation and cytokine production.

After obtaining the 20 characteristic urine proteins, the applicant trains a machine learning model based on the relative expression amounts of the 20 urine proteins, and classifies covid-19 weight types based on the trained model.

Specifically, relative expression amounts of urine proteins of urine of different light and heavy covid-19 subjects are used as training samples to train a machine learning model, wherein the urine proteins are as follows: VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84 and BTNL3, the predicted value characterizing the covid-19 lightness and heaviness degree of the subject.

The model used by the machine learning model was constructed using the R software package (version 4.6.14) of the random forest. Specifically, the relative expression of the 20 proteins in 39 light patients and 11 heavy patients is used as raw data input into a model, and the main parameters of model construction are set as follows: ntree is 1000, nodesize is 1, others use default parameters.

The training samples are: relative quantitative data of urine protein of 39 light patients and 11 heavy patients are obtained, and the light and heavy types of the corresponding specimens are marked.

Training process: the relative expression of the 20 proteins in 39 light patients and 11 heavy patients is used as the raw data of an input model to be input into the model structure for training.

The scheme verifies that the parting model performs well when the covid-19 is predicted to be light and heavy by the following experiments.

Typing experiment 1:

experimental samples: urine samples from 13 covid-19 patients were collected using conventional collection methods, and were obtained from taizhou hospital, zhejiang. In this scheme, the conventional acquisition means refers to: the patient urinates by himself and then is collected by a urine collecting tube.

And (3) measuring the urine sample by using a tmt-labeled urine proteome analysis method to obtain the relative expression amount of the urine protein of the urine sample, wherein the urine protein is VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84 and BTNL 3. In particular, the assay for urine samples is described above in relation to the sample preparation and nanoliter liquid-high resolution mass spectrometry steps.

The results obtained by inputting the relative expression amount of urine protein into the typing model are shown in fig. 3, and it can be seen that the area under the curve (AUC) is 0.89, and the accuracy of the predicted value as a whole is 0.69. The curve can be used for evaluating the effectiveness of the model of the scheme, and the stability and the performance of the model can be evaluated by using the model obtained by screening 50 training queue samples and using the related parameter (AUC) of the curve. The established model is then applied to a subsequent independent queue (13-sample queue) for typing prediction. The sensitivity and specificity of the typing results can also be demonstrated by the correlation parameter (AUC) of this curve. If the training queue modeling and the independent queue evaluation have higher related parameters, the model effectiveness is better, and the method is scientific.

And (4) analyzing results: the relative expression level of urine protein in 13 covid-19 patients is shown in FIG. 4, and the predicted values and disease states are shown in Table 1.

Table one: predictive value and Condition in 13 covid-19 patients

Patient numbering	Mild case of disease	Severe illness
			U10_10	0.972	0.028
U10_11	0.189	0.811
			U10_15	0.13	0.87
U7_13	0.296	0.704
			U7_2	1	0
U7_8	1	0
			U8_15	0.828	0.172
U8_3	0.872	0.128
			U8_6	0.972	0.028
U8_9	0.235	0.765
			U9_11	1	0
U9_3	0.668	0.332
			U9_7	0.999	0.001

As can be seen from the results, 6 non-critically ill patients (U9_7, U7_8, U9_11, U8_3, U10_10, U7_2) and 3 critically ill patients (U10_11, U10_15, U8_9) were correctly classified.

Of the 4 misclassified samples, severe case U8-6 was a 36 year old female, and urine was collected 3 days after diagnosis of a severe patient (severity index: oxygenation index < 300). However, at 2 days after diagnosis of severe disease, the oxygenation index rose to 324, indicating that sampling was occurring during the remission phase. In fact, a CT examination three days after urine sampling also showed that her two lungs showed a decreasing trend in their multispot shadows. Therefore, the urine sample collection of the patient is in the improvement stage, which results in the misclassification of the case, but it also reflects the typing model from the side to achieve a certain predictive typing effect. Example 2 (U7-13) is non-severe and was misdiagnosed as severe, with a history of actual type 2 diabetes of 2 years, and may affect model evaluation. The other two cases, U9-3 and U8-15, were severe in men and were misdiagnosed as non-severe, for unknown reasons.

In addition, supplementary control experiment 2:

urine samples of 13 covid-19 patients identical to the classification experiment 1 were selected and tested by tmt-labeled uroproteomics analysis method to obtain the relative expression amount of urine protein in the urine samples, and the only difference is that another 20 proteins were randomly selected for modeling: GASK1B, LIPA, PLAUR, RPL19, LSM2, GALM, FAM25A, RRAS, C1RL, FOLH1, NUCKS1, NPM1, S100A6, FAT4, EPS8L2, C5orf15, SIAE, CRK, SEMG1, SMPDL 3B.

The machine learning model is: the same model structure as experiment 1 was chosen, with the parameters set to ntree 1000 and nodesize 1.

The training samples are: urine protein relative quantitative data of 39 light patients and 11 heavy patients, and marking the light and heavy types of corresponding specimens

The result of inputting the relative expression amount of urine protein into the typing model 2 at this time is shown in fig. 5, in which the area under the curve (AUC) is 0.762 and the accuracy is 0.62.

The predicted values and disease status of 13 covid-19 patients are shown in Table 2.

Table 2: predictive value and condition of 13 covid-19 patients in control experiment 2

As can be seen from the results, almost all patients were summarized as light-weight, without discriminative power.

Compared with the control experiment 2 and the classification experiment 1, the method can obviously sense that the prediction result obtained by predicting the relative expression quantity of the 20 kinds of urine proteins selected by the scheme is more accurate.

The present invention is not limited to the above-mentioned preferred embodiments, and any other products in various forms can be obtained by anyone in the light of the present invention, but any changes in the shape or structure thereof, which have the same or similar technical solutions as those of the present application, fall within the protection scope of the present invention.

Claims

1. The application of the characteristic urine protein as a urine detection target in preparing a reagent kit for typing a covid-19 light-weight grade of a subject is characterized in that the reagent kit contains a reagent for detecting the relative expression quantity of the characteristic urine protein; detecting the relative expression quantity of the characteristic urine protein in the urine of the subject by using the kit, wherein the relative expression quantity of the characteristic urine protein is related to the covid-19 typing result of the subject, and the characteristic urine protein is as follows: VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84, and BTNL 3.

2. The use of claim 1, wherein the relative amount of expression of characteristic urine proteins is correlated with subject covid-19 typing results: inputting the relative expression quantity of the characteristic urine protein into a typing model, and if the output predicted value of the typing model is lower than a set threshold value, predicting that the tested person is covid-19 heavy.

3. The use of claim 1, wherein the subject's covid-19 typing results are correlated with the subject's renal tissue monitoring, and the predictive value of the typing model is inversely correlated with the degree of renal tissue damage.

4. The application of the characteristic urine protein in preparing a reagent for predicting the covid-19 typing result is characterized in that the relative expression quantity of the characteristic urine protein in the urine of a subject is detected and input into a typing model to predict the covid-19 typing result of the subject, wherein the characteristic urine protein is as follows: VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84, and BTNL 3.

5. A construction method of a typing model for typing covid-19 light and heavy grades is characterized in that relative expression quantities of characteristic urine proteins of urine of different light and heavy grades of covid-19 subjects are used as training samples to train a machine learning model to obtain the typing model, wherein the characteristic urine proteins are as follows: VPS36, CEL, FREM2, TNR, PTGFRN, MLEC, ADGRL1, OSTN, ICOSLG, NPR3, MELTF, PRSS2, SERPINI1, MADCAM1, CDH22, CDH19, TRHDE, SPOCK1, CD84 and BTNL3, and the predictive value of the typing model characterizes covid-19 lightness and heaviness of the subject.

6. The method for constructing the parting model for parting covid-19 light and heavy grades as claimed in claim 5, characterized in that the machine learning model is constructed by a random forest algorithm.