CN112802544A

CN112802544A - System for predicting urinary calculus based on machine learning and metabonomics

Info

Publication number: CN112802544A
Application number: CN202110273193.8A
Authority: CN
Inventors: 高飞; 海云; 谷云云; 覃岚芯; 罗示齐
Original assignee: Hangzhou Durbrain Medical Inspection Laboratory Co ltd
Current assignee: Hangzhou Durbrain Medical Inspection Laboratory Co ltd
Priority date: 2021-03-12
Filing date: 2021-03-12
Publication date: 2021-05-14

Abstract

The invention discloses a system for predicting urinary calculus based on machine learning and metabonomics, and belongs to the field of medical diagnosis based on new-generation information technology. The system comprises a database device for storing the data composed of group characteristics; a data input device for receiving input of subject characteristic data; and a prediction device for establishing a prediction model by using a machine learning method based on the characteristic data in the database and predicting whether the subject is at risk of urinary calculus occurrence or not by using the prediction model based on the characteristic data of the subject, wherein the characteristic data comprises urine isocitric acid content and urine citric acid content. The system can accurately predict the urinary calculus occurrence risk of the testee based on the characteristic data of the testee, such as the isocitric acid content and the citric acid content in urine, and based on machine learning, so as to perform early intervention, and has important clinical application value.

Description

System for predicting urinary calculus based on machine learning and metabonomics

Technical Field

The invention belongs to the field of medical diagnosis based on a new generation information technology, and particularly relates to a system for predicting urinary calculus based on machine learning and metabonomics.

Background

Urinary calculus (Nephrolithiasis) is caused by abnormal accumulation of crystalline substances in the urinary system, is a common disease of the urinary system, and can cause urinary tract infection, abdominal pain, hydronephrosis, renal failure and the like seriously. Urinary calculus has high morbidity, the morbidity in southern China is as high as 5-10%, the urinary calculus is easy to relapse, the life quality of patients after operation is low, the burden on families and society of patients is huge, many patients can find the calculus after symptoms appear, the calculus risk cannot be evaluated in time in the early stage, and an effective detection mode is not available for the evaluation of the treatment effect after the calculus operation. The calculus recurrence rate is extremely high, so that the establishment of a suitable risk assessment model and a suitable detection means has important practical significance in assessing the treatment effect of a calculus patient and predicting the calculus recurrence or occurrence risk of the patient.

The traditional method for determining calculus is usually medical image such as B-ultrasonic, X-ray or CT detection, and calculus must be generated already during patient examination, so that the method cannot be used for early calculus prediction. There is also some hysteresis in assessing the efficacy of treatment for patients with stones.

In recent years, the appearance of biomarkers, such as oxalic acid, uric acid and calcium in urine, makes it possible to predict the occurrence of calculi and to evaluate the therapeutic effect of calculi patients. However, the human body has an integral metabolic system, calculus production is comprehensively influenced by various factors, and the existing biomarkers lack integral research data.

Metabolomics is the comprehensive assessment of the condition of an organism by studying the wide range of metabolite changes in the organism. Metabolomics has found applications in a number of fields of human disease research. However, in the aspect of urinary calculus, such comprehensive data is not available, so that the research on urinary calculus by a metabonomics method is more realistic.

On the other hand, clinical urinary stone diagnosis based on other phenotypes is also of great significance. However, no technique or method has been reported to combine biomarkers with clinical phenotypes for urinary stone prediction.

Disclosure of Invention

In order to solve at least one of the above technical problems, the technical solution adopted by the present invention is as follows:

the invention provides a system for predicting urinary calculus occurrence risk of a subject based on machine learning, which comprises:

database means for storing a database composed of group feature data;

a data input device for receiving input of subject characteristic data;

urinary calculus occurrence risk prediction means connected to the database means and the data input means, respectively, for establishing a prediction model by a machine learning method based on the characteristic data in the database, and predicting whether the subject is at risk of occurrence of urinary calculus based on the subject characteristic data using the prediction model,

wherein the characteristic data comprises urine isocitric acid content and urine citric acid content.

Further, the characteristic data further includes at least one selected from the group consisting of sex, body weight, urine volume, urinary oxalic acid content, urinary calcium content, urinary creatinine content, uric acid content, and age.

In some embodiments of the invention, the characteristic data includes urine isocitric acid content, urine citric acid content and urine volume.

In other embodiments of the present invention, the characteristic data includes urine isocitric acid content, urine citric acid content and age.

In still further embodiments of the present invention, the characteristic data includes urine isocitric acid content, urine citric acid content, urine volume, and age.

In some embodiments of the invention, the population characteristic data refers to corresponding characteristic data of a large number of individuals, such as more than 20, more than 50, more than 100, more than 500 or more individuals, and data of whether a urolithiasis has occurred over a period of time. In some preferred embodiments of the invention, the certain period of time is 1 year.

Further, the urolith prediction device is further configured to input the subject characteristic data into the database device, update the database, and create a new database.

As such, in some embodiments of the invention, the urolith prediction device generates a new prediction model based on a new database.

In the invention, the machine learning method is a random forest method or a single factor analysis method.

In the present invention, the system further comprises a prediction result output device connected to the urolith prediction device.

In some embodiments of the invention, the urine is 24h urine.

In the present invention, the subject is a human.

In some embodiments of the present invention, the urine isocitric acid content, urine citric acid content and urine oxalic acid content are measured by using a high performance liquid chromatography-tandem mass spectrometry method, which comprises the following steps:

s1, derivatizing the urine sample by using a derivatizing reagent;

and S2, detecting the derivatized urine sample by using high performance liquid chromatography-tandem mass spectrometry.

In some embodiments of the present invention, the step of derivatizing the urine sample with a derivatizing reagent in step S1 is specifically: taking 10-20 mu L of biological sample, adding 10-50 mu L of isotope labeled isocitric acid and oxalic acid internal standard, drying at 60 ℃ in nitrogen, adding 100-300 mu L of 3mol/L hydrochloric acid n-butyl alcohol solution, swirling for 3min, shaking at 60 ℃ for 20min, centrifuging for 3min, drying at 60 ℃ in nitrogen, and adding 100-300 mu L of methanol for redissolution.

In some embodiments of the present invention, step S2 is preceded by the further step of:

after derivatization, drying the sample by nitrogen, respectively adding 0.5-1.5mL of 0.3% ammonia water and ethyl acetate, carrying out ultrasonic treatment for 15min, shaking for 15min, centrifuging at 15000rpm for 5min, taking 800 mu L of supernatant, drying by nitrogen at 60-80 ℃, adding 50-500 mu L of methanol for redissolution, and detecting.

In some embodiments of the present invention, the step S2 of detecting the derivatized biological sample by high performance liquid chromatography-tandem mass spectrometry specifically comprises:

chromatographic column conditions: ACE Excel-2C 18-PFP column (100X 2.1 mm, 2.6 μm) with column temperature of 35 deg.C;

elution conditions: the mobile phase A is 0.1% formic acid-5 mM ammonium acetate aqueous solution, and the mobile phase B is 0.1% formic acid-5 mM ammonium acetate methanol solution; flow rate 0.1-0.5mL/min, isocratic elution, 10-90% B mobile phase B, as detailed in the following table.

In some embodiments of the invention, the system is a computer system.

The invention has the advantages of

Compared with the prior art, the invention has the following beneficial effects:

by using the system provided by the invention, the urinary calculus occurrence risk of the testee can be accurately predicted by using a machine learning method based on metabonomics characteristic data of the testee, namely the citric acid content and the isocitric acid content in urine, so that early intervention is carried out, and the system has an important clinical application value.

Drawings

Figure 1 shows the structural and biochemical reactions of citric acid with isocitric acid.

FIG. 2 shows a peak plot for biomarker detection based on LC-MS/MS.

Figure 3 shows the relationship between isocitric acid and oxalic acid content in a 24h urine sample from a subject.

Figure 4 shows the parameter importance ranking fitted by the random forest method.

Figure 5 shows a schematic diagram of a system for predicting the risk of occurrence of a urolithiasis based on machine learning according to embodiment 4 of the present invention.

Figure 6 shows an ROC curve for predicting stone occurrence based on a system for predicting urinary stone occurrence risk based on machine learning in example 4 of the present invention. Wherein the solid line is the training set (Train) and the dotted line is the Test set (Test).

Detailed Description

In order to make the technical problems, technical solutions and advantageous effects solved by the present invention more apparent, the present invention is further described in detail below with reference to the following embodiments.

Examples

The following examples are used herein to demonstrate preferred embodiments of the invention. It will be appreciated by those of skill in the art that the techniques disclosed in the examples which follow represent techniques discovered by the inventor to function in the invention, and thus can be considered to constitute preferred modes for its practice. Those of skill in the art should, in light of the present disclosure, appreciate that many changes can be made in the specific embodiments which are disclosed and still obtain a like or similar result without departing from the spirit or scope of the invention.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs and the disclosures and references cited herein and the materials to which they refer are incorporated by reference.

Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments of the invention described herein. Such equivalents are intended to be encompassed by the following claims.

The experimental procedures in the following examples are conventional unless otherwise specified. The instruments used in the following examples are, unless otherwise specified, laboratory-standard instruments; the test materials used in the following examples were purchased from a conventional biochemical reagent store unless otherwise specified.

Example 124 h separation and detection of Isocitric acid in urine samples

Isocitric acid and citric acid are endogenous substances of isomers (figure 1), and in the conventional detection method, the separation detection of citric acid and isocitric acid is difficult to realize.

Preparing a standard solution.

(1) Preparing an internal standard solution: accurately weighing citric acid-D by using analytical balance₄And dissolving the internal standard by pure water to prepare internal standard mother liquor with the concentration of 10 mg/mL.

(2) Preparing a standard substance: accurately weighing citric acid and isocitric acid by an analytical balance, and preparing the standard substances into standard substance mother liquor with the concentration of 10mg/mL and 50mg/mL by using pure water respectively.

(3) Internal standard stock solution: and sucking 100 mu L of internal standard mother liquor, using pure water to fix the volume to 1mL, and diluting into an internal standard stock solution.

(4) Standard stock solutions: sequentially sucking 500 mu L and 300 mu L of citric acid and isocitric acid standard mother liquor, fixing the volume to 1mL by using pure water, and diluting into a standard product stock solution.

(5) Preparing artificial urine: adding 180mg of urea, 5mg of uric acid and 110mg of NaCl into 10mL of ultrapure water, carrying out ultrasonic treatment for 10min, uniformly mixing for 3min, standing for 5min, and taking supernatant solution for later use.

(6) Standard curve working fluid: the standard substance stock solution is diluted by using artificial urine to obtain a series of standard curve working solutions (W1-W7) with different concentrations, and the standard curve working solutions are used for preparing a standard curve. The standard curve working solution preparation process is as follows:

(II) sample pretreatment

Centrifuging the urine: taking at least 4mL of urine sample to be detected (24 h urine of a subject), centrifuging for 15min at a centrifugation speed of 3000rpm, separating to obtain supernatant urine, and storing at-80 ℃ for later use.

(III) derivatization

And (3) adding 20 mu L of internal standard into 20 mu L of urine sample to be detected/standard curve working solution, drying at 60 ℃ by nitrogen, adding 100 mu L of 3mol/L hydrochloric acid n-butyl alcohol solution, swirling for 3min, shaking at 60 ℃ for 20min, centrifuging for 3min, drying at 60 ℃ by nitrogen, and extracting.

(IV) extracting: after derivatization, drying the sample by nitrogen, respectively adding 1mL of 0.3% ammonia water and ethyl acetate, carrying out ultrasonic treatment for 15min, shaking for 15min, centrifuging at 15000rpm for 5min, taking 500 mu L of supernatant, drying by nitrogen at 60 ℃, adding 100 mu L of methanol for redissolving, and detecting.

(V) detection of urine sample to be detected/standard curve working solution

Chromatographic column conditions: ACE Excel-2C 18-PFP column (100X 2.1 mm, 2.6 μm) at 35 ℃.

Elution conditions: fluidity A is 0.1% formic acid-5 mM ammonium acetate aqueous solution, fluidity B is 0.1% formic acid-5 mM ammonium acetate methanol solution; flow-0.3 mL/min, isocratic elution, -90% B mobile phase B, as specified in the following table:

the detection result is shown in fig. 2, and the result shows that the citric acid and the isocitric acid realize baseline separation, and the retention time of the isocitric acid is 2.5-2.8 min; the retention time of citric acid is 2.85-3.25 min.

Example 2 biomarker assay

Preparing a standard solution.

(1) Preparing an internal standard solution: accurately weighing oxalic acid by analytical balance¹³C₂Citric acid-D_4、cystine-D₄The weighed mass is 10mg, 100mg and 10mg in turn, each internal standard is dissolved by pure water respectively, and internal standard mother liquor with the concentration of 10mg/mL, 10mg/mL and 10mg/mL in turn is prepared.

(2) Preparing a standard substance: accurately weighing oxalic acid, citric acid, isocitric acid and cystine by an analytical balance, wherein the weighed mass is 20mg, 500mg, 50mg and 30mg in sequence, and preparing the standard mother solution with the concentrations of 20mg/mL, 500mg/mL, 50mg/mL and 30mg/mL by pure water respectively for each standard.

(3) Mixing internal standard stock solution: sequentially absorb 100 mu L of oxalic acid-¹³

C

₂100. mu.L of citric acid-D₄And 100. mu.L cystine-D₄And (4) diluting the internal standard mother liquor to 1mL by using pure water to obtain a mixed internal standard stock solution.

(4) Mixing standard stock solution: and sucking 500 mu L of oxalic acid mother liquor, 500 mu L of cystine mother liquor, 500 mu L of isocitric acid mother liquor and 500 mu L of citric acid standard product mother liquor, fixing the volume to 1mL by using pure water, and diluting into mixed standard product stock solution.

(5) Standard curve working fluid: the standard substance stock solution is diluted and mixed by using artificial urine to obtain a series of standard curve working solutions (W1-W7) with different concentrations, and the standard curve working solutions are used for preparing a standard curve. The standard working solution formulation procedure is as follows:

(II) sample pretreatment

(III) derivatization

(V) detection of sample to be detected

Elution conditions: fluidity A is 0.1% formic acid-5 mM ammonium acetate aqueous solution, fluidity B is 0.1% formic acid-5 mM ammonium acetate methanol solution; flow rate 0.5mL/min, isocratic elution, 90% B mobile phase B, as specified in the table below.

The test results of the urine sample are shown in FIG. 2.

Detecting the working solution of the standard curve; and fitting to obtain a standard curve equation through the concentration and the peak area of the standard curve, and calculating the concentrations (contents) of the oxalic acid, the citric acid, the isocitric acid and the cystine in the urine sample to be detected through the peak area of the urine sample to be detected.

The inventors performed the test on urine samples (24 h urine) of 6 subjects using the method of this example, and the results are shown in the following table:

up to the normal range; ↓ denotes lower than normal range.

The inventor tests a large number of samples, and statistically discovers that the isocitric acid content and the oxalic acid content in urine samples of subjects have a significant positive correlation, and the Pearson linear correlation coefficient r =0.397 (FIG. 3) indicates that the isocitric acid effect is inconsistent with the citric acid, which may promote the formation of oxalic acid. Meanwhile, the method also means that the citric acid and the isocitric acid are separated and simultaneously detected, and the method has great clinical significance for diagnosing and predicting the urinary calculus.

Example 3 model for predicting urolithiasis risk based on machine learning

In order to better predict the occurrence and recurrence of calculus, the inventor collects 24h urine of healthy people and 24h urine of calculus patients (calculus occurs in the last year), measures the contents of the biomarkers oxalic acid, citric acid, cystine and isocitric acid by the LC-MS/MS method established in example 2, obtains indexes such as potassium in urine, calcium in urine, sodium in urine and the like by a biochemical analyzer, obtains other clinical phenotype information (such as age, sex, weight and the like), and obtains part of the subject information as follows:

the inventor screens important prediction indexes (Predictive Features) in a machine language learning (machine learning) mode, and finally compares and determines a final prediction model and parameters through an ROC curve.

Model one: the top 5 important Features (Features) were found by the Random forest method (Random forces) (fig. 4), and the variable with p value <0.05 was further found by the method of Logistic Regression: citric acid content, isocitric acid content, urine volume and age.

Model two: the characteristic p value <0.05 (Features) was found by the One way ANOVA method: citric acid content, isocitric acid content, age, urine volume, urine magnesium, body weight, pH, cystine content, as shown in the table below. For these variables, a variable with p value <0.05 was further found for the method by Logistic Regression: citric acid content, isocitric acid content, age and urine volume.

Through the first model and the second model, it can be seen that the important prediction indexes (Predictive Features) screened by the two models are consistent. Therefore, the inventor selects a prediction model of citric acid content + isocitric acid content + age + urine amount as a final model for predicting the occurrence risk of calculus by means of Logistic regression.

Example 4 System for predicting urolithiasis risk based on machine learning

Based on example 3, the inventors set up a system as shown in fig. 5, which comprises: a database device 1 for storing a database composed of group feature data; a data input device 2 for receiving input of subject characteristic data; a urinary calculus occurrence risk prediction device 3 connected to the database device 1 and the data input device 2, respectively, for establishing a prediction model by a machine learning method based on the characteristic data in the database device 1 and predicting whether the subject has a risk of occurrence of urinary calculus based on the subject characteristic data by using the prediction model, and a prediction result output device 4 connected to the urinary calculus occurrence risk prediction device 3 for outputting the risk of occurrence of urinary calculus of the subject.

Wherein the characteristic data comprises the isocitric acid content in 24h urine, the citric acid content in 24h urine, the urine volume in 24h and the age of 100 individuals. The database also included the results of whether 100 individuals had urolithiasis within 1 year. Some results are shown in example 3.

The inventors validated this system by using further random creation of training sets and test sets (figure 6). Where the training set AUC was 0.868 (solid line) and the test set AUC was 0.785 (dashed line).

The above results show that the system established by the inventor for predicting the risk of urinary calculus based on machine learning has very high accuracy.

During use, the urinary stone prediction device 3 can input the characteristic data of the subject and whether the urinary stone occurs within 1 year into the database device 1, and update the database, thereby forming a new database. At the next prediction, the urinary stone prediction device 3 may generate a new prediction model based on the new database. Along with the accumulation of data, the database is continuously perfected and upgraded, and the accuracy of the prediction model is continuously improved.

All documents referred to herein are incorporated by reference into this application as if each were individually incorporated by reference. Furthermore, it should be understood that various changes and modifications of the present invention can be made by those skilled in the art after reading the above teachings of the present invention, and these equivalents also fall within the scope of the present invention as defined by the appended claims.

Claims

1. A system for predicting urolithiasis based on machine learning and metabolomics, comprising:

database means for storing a database composed of group feature data;

a data input device for receiving input of subject characteristic data;

urinary calculus occurrence risk prediction means connected to the database means and the data input means, respectively, for establishing a prediction model by a machine learning method based on the feature data in the database, and predicting whether the subject is at risk of occurrence of urinary calculus based on the feature data using the prediction model,

2. The system of claim 1, wherein the characteristic data further comprises at least one selected from the group consisting of gender, weight, urine volume, urinary oxalic acid content, urinary calcium content, urinary creatinine content, uric acid content, and age.

3. The system of claim 1 wherein said urolithiasis prediction device is further configured to input said subject characteristic data into said database device, update said database, and create a new database.

4. The system of claim 3 wherein said urolithiasis prediction device generates a new prediction model based on a new database.

5. The system of claim 1, wherein the machine learning method is a random forest method or a single factor analysis method.

6. The system of claim 1, further comprising a prediction output device coupled to the urolith prediction device.

7. The system of any one of claims 1-6, wherein the urine is 24h urine from a subject.