CN110780070A

CN110780070A - Plasma protein molecule for detecting cancer chemotherapy sensitivity, application and kit

Info

Publication number: CN110780070A
Application number: CN201910924077.0A
Authority: CN
Inventors: 梁廷波; 白雪莉; 章琦; 楼煜; 叶茅
Original assignee: Zhejiang University ZJU
Current assignee: Zhejiang University ZJU
Priority date: 2019-09-27
Filing date: 2019-09-27
Publication date: 2020-02-11
Anticipated expiration: 2039-09-27
Also published as: CN110780070B

Abstract

The invention discloses a plasma protein molecule for detecting cancer chemotherapy sensitivity, application and a kit. A plasma protein molecule for detecting cancer chemotherapy sensitivity comprises at least one of human plasma proteins GSN, APOA4, IGHG1, immunologlobulin mu heavy chain and FCN 2. The research of the invention finds that the concentration of the human plasma protein has obvious difference in pancreatic cancer patients with different chemotherapy responses, and the area under the ROC curve for predicting the pancreatic cancer chemotherapy response can reach 0.5550-0.7275 when the human plasma protein is used as a tumor marker alone; the area under the ROC curve of the five human plasma protein concentration combinations used as tumor markers for predicting the pancreatic cancer chemotherapy response in combination with the age of a patient can reach 0.915, and the human plasma protein concentration combinations can be used as tumor markers for predicting the tumor chemotherapy sensitivity, so that the effective screening of chemotherapy-sensitive people is realized, and the clinical benefit is greatly improved.

Description

Plasma protein molecule for detecting cancer chemotherapy sensitivity, application and kit

Technical Field

The invention relates to the technical field of biotechnology detection, in particular to a plasma protein molecule for detecting cancer chemotherapy sensitivity, application and a kit.

Background

Tumors are one of the important diseases threatening human life and health. Pancreatic cancer is the most malignant tumor, and is called the king of cancer. The occult onset of the cancer is such that many patients often lose the chance of surgical operation during diagnosis, and the poor response rate of chemotherapy greatly limits the prognosis of pancreatic cancer patients.

With the research of multiple researches in recent years, some chemotherapy schemes such as FOLFIRINOX, nano albumin paclitaxel and gemcitabine are gradually separated out, and become a main chemotherapy strategy for treating pancreatic cancer. Unfortunately, however, these regimens remain very limited in their effectiveness in the pancreatic cancer population. Therefore, the search for effective tumor markers helps to accurately identify chemotherapy-benefited people, and has important significance for improving the prognosis of pancreatic cancer patients. CA19-9 is one of the most commonly used tumor markers in clinic, and has good sensitivity and specificity for pancreatic cancer diagnosis. Based on the expression pattern of CA19-9, it was found that the FOLFIRINOX regimen resulted in a higher objective sustained release rate (ORR, 44.0% vs. 22.9%) for patients with a more than 20% decrease in CA 19-9. However, CA19-9 is not expressed in all pancreatic cancer patients, nor is this strategy able to predict patient response prior to chemotherapy. To solve this problem, some studies have searched for markers of tumor circulating DNA and serum proteins, and found that biomarkers such as ctdna (kras), CEA, and sCD40L can predict the sensitivity of chemotherapeutic drugs to some extent. However, the sensitivity and specificity of prediction of these biomarkers, either alone or in combination, have been difficult to meet clinical needs. Therefore, the development of a new tumor marker has good application prospect in predicting drug sensitivity before chemotherapy.

With the development and maturation of technology, mass spectrometry-based proteomics strategies are gradually expanding in clinical detection and basic research. The technical characteristics of high flux, high sensitivity and high accuracy bring great convenience to the biomedical field, and the large-scale screening of the biomarkers becomes possible. Therefore, the strategy of utilizing omics big data is used for screening new markers and constructing a prediction model of multiple markers, or the accuracy of chemotherapy prediction can be improved to a new level.

Disclosure of Invention

The research of the invention finds that the plasma protein concentrations of free proteins GSN, APOA4, IGHG1, immunoglobinum muaviy chain (IgM heavy chain), FCN2 extracted from human peripheral blood show significant difference in pancreatic cancer patients with different chemotherapy responses, and further analysis can find that the high expression of serum GSN, APOA4, immunoglobinum heavy chain, FCN2 and the low expression of IGHG1 are related to the sensitivity of tumors to the FOLFIRINOX scheme. The protein molecule is used for constructing a pancreatic cancer chemotherapy response prediction model, the area under the working characteristic (ROC) curve of a subject can reach 0.88, and the area under the ROC curve for predicting pancreatic cancer chemotherapy response by combining the age of a patient can reach 0.915. Therefore, the prediction model constructed by combining the plasma protein marker molecules with the clinical information of the patient can be used as a tumor marker for predicting tumor chemotherapy response, and further helps to clinically screen chemotherapy-sensitive people.

A plasma protein molecule for detecting cancer chemotherapy sensitivity comprises at least one of human plasma proteins GSN, APOA4, IGHG1, immunologlobulin mu heavy chain and FCN 2.

Preferably, the plasma protein molecule comprises one of the following protein combinations: GAI, GAF, GAM, GIF, GIM, GFM, AIF, AFM, IFM, AIM, wherein G: GSN, A: APOA4, I: IGHG1, M: immunoglobin mu heavychain, F: FCN 2.

More preferably, the plasma protein molecules comprise human plasma proteins GSN, APOA4, IGHG1, immunologlulin mu heavy chain and FCN 2. The result is more accurate when the five human plasma proteins are combined together.

The invention also provides application of the plasma protein molecule in serving as a target for detecting cancer chemotherapy sensitivity. The application, the cancer type is pancreatic cancer. The chemotherapy is FOLFIRINOX.

The application comprises the steps of detecting the concentrations of human plasma proteins GSN, APOA4, IGHG1, immunologlulin Muheavy chain and FCN2 in plasma, and inputting the 5 kinds of human plasma protein concentration data into a pre-generated random forest model for prediction and evaluation of the chemotherapy response of a patient.

The construction method of the pre-generated random forest model comprises the following steps:

taking concentration data of human plasma proteins GSN, APOA4, IGHG1, Immunoglobulin mu heavy chain and FCN2 in the plasma of patients with known FOLFIRINOX regimen sensitivity and resistance,

setting the seed number as 2019, randomly grouping according to the ratio of 2: 1 to construct a training set and a verification set, and importing concentration matrixes of 5 proteins, treatment response conditions and corresponding clinical information: and (3) constructing a model by using a random forest algorithm, repeating the operation for 100 times to obtain an ROC working curve for average prediction of the model, setting the number of trees to be 1000, setting the minimum observed data number of the peripheral nodes to be 1: 5, and fitting by using a anger function to obtain a final model.

The present invention also provides a kit for detecting cancer chemotherapy sensitivity, comprising:

(1) reagents for extracting proteins from plasma;

(2) the reagent is used for detecting the concentrations of human plasma proteins GSN, APOA4, IGHG1, immunologlobulin muheavy chain and FCN 2.

The research of the invention finds that the concentrations of human plasma proteins GSN, APOA4, IGHG1, immunologlulin muheavychain and FCN2 are obviously different in pancreatic cancer patients responding to different chemotherapies, and the area under the ROC curve for predicting the pancreatic cancer chemotherapies response can reach 0.5550-0.7275 when the human plasma proteins GSN, APOA4, IGHG1, immunologlulin muheavychain and FCN2 are used as tumor markers independently; the area under the ROC curve for predicting the pancreatic cancer chemotherapy response by using the five human plasma protein concentration combinations as tumor markers in combination with the patient age can reach 0.915. Therefore, based on the combination of plasma protein molecules GSN, APOA4, IGHG1, immunolobulin muheavy chain and FCN2 (including any combination of the five proteins and any combination of all parts of the five proteins) and clinical information of a patient, the tumor marker for predicting tumor chemotherapy sensitivity can be used as a tumor marker for assisting with a random forest model constructed in the early stage, so that the effective screening of chemotherapy-sensitive people is realized, and the clinical benefit is greatly improved.

Drawings

FIG. 1 is a graph of plasma protein concentrations versus results for chemotherapy resistant and sensitive pancreatic cancer patients.

FIG. 2 is a diagram of the decision tree results of one of the models extracted during model building in example 3.

FIG. 3 is a graph of the results of sensitivity and specificity analysis of plasma protein combinations in the prediction of pancreatic cancer chemotherapy response.

FIG. 4 is a graph of the results of a sensitivity and specificity analysis of the single protein marker model in the prediction of pancreatic cancer chemotherapy response.

FIG. 5 is a graph of the results of sensitivity and specificity analysis in response prediction of pancreatic cancer chemotherapy using a traditional three-protein marker combination model based on binary Logistic regression analysis.

Detailed Description

Sample source: plasma samples from pancreatic cancer patients were obtained from the first hospital affiliated with the university of Zhejiang medical college.

Ethical examination and approval: ethical review by ethical review committee of scientific research in first hospital affiliated to Zhejiang university medical college, lot number: (2019) research review quick review No. (622).

Example 1

Plasma proteins were extracted from peripheral blood samples. The method comprises the following steps:

peripheral blood was collected from human using an anticoagulation blood collection tube, centrifuged (3000g, 4 ℃ C., 20 minutes) to remove blood cells, and the supernatant was frozen at-80 ℃ in a refrigerator. During detection, the supernatant is taken out and frozen, and then the supernatant is thawed at 4 ℃. Centrifugation (10000g, 60 min) was carried out to remove large cell debris and the like. Mu.l of the supernatant was added to 200. mu.l of a protein extract (purchased from Biognosys, product number of kit: Ki-3013) and mixed well, followed by standing at 4 ℃ for 5 minutes. After centrifugation (14000g, 4 ℃, 10 minutes), the supernatant was removed to obtain a plasma protein sample.

Example 2

Plasma protein samples were quantitated using the method steps of a protein quantitation kit (purchased from Thermo Fisher, product number: 23225). The protein was subjected to enzymatic hydrolysis using protease (purchased from Biognosys, Inc., kit product No. Ki-3013) to obtain a peptide fragment which was used for processing. And according to the quantitative result, the same amount of nuclide-labeled standard peptide fragment (purchased from Biognosys company, product number: Ki-3019) is added to the polypeptide sample with the same content to prepare the peptide fragment sample for being installed on a computer.

Example 3

The sample in example 2 was subjected to mass spectrometric detection to obtain a quantitative result of the target protein. And calculating the plasma concentration of the target protein according to the absolute quantitative result and the volume of the upper computer sample, and introducing the obtained concentration into a random forest model constructed in the early stage so as to predict the chemotherapy response condition of the patient.

The test relates to 40 plasma samples of pancreatic cancer chemotherapy patients, wherein the FOLFIRINOX scheme is sensitive to 25 patients and the drug-resistant patients are 15 patients, and the results show that the levels of GSN, APOA4, IGHG1, immunologlulin mu heavy chain and FCN2 protein in the peripheral blood of the patients with different chemotherapy responses are remarkably different (figure 1).

Setting the seed number as 2019, randomly grouping according to the ratio of 2: 1 to construct a training set and a verification set, and importing concentration matrixes of 5 proteins, treatment response conditions and corresponding clinical information: and (3) constructing a model by using a random forest algorithm according to the sex, the age, the tumor marker and the tumor size and position, and repeating the operation for 100 times to obtain an ROC working curve for average prediction of the model. Setting the number of trees as 1000, the minimum observed data number of the peripheral nodes as 1: 5, and fitting by using a anger function to obtain a final model. The decision tree of the extraction model shows that the 5 serum proteins can be used for efficiently predicting the pancreatic cancer of the sample set, and the accuracy is as high as 95%. Based on this decision tree, it could be found that high expression of serum GSN, APOA4, immunolobulin mu heavy chain, FCN2 and low expression of IGHG1 correlated with tumor sensitivity to the FOLFIRINOX regimen (fig. 2).

Example 4

Based on the analysis of the random forest model on the verification queue, the ROC curve shows that the prediction model based on the combination of plasma protein molecules GSN, APOA4, IGHG1, immunolobulin mu heavy chain and FCN2 and the age of the patient can effectively predict the chemotherapy response condition of the pancreatic cancer patient, and the area under the ROC curve can reach 0.915 (figure 3). In particular, sensitivity can reach more than 80% while ensuring 100% diagnostic specificity. The prediction efficiency is significantly better than that of the traditional three-protein marker combination model (fig. 5 and table 1) established based on the binary Logistic regression analysis and in the presence of the single-protein marker model (the result is shown in fig. 4 by using the wilson/brown model). The latter uses a single queue for training and verification, so that the risk of overfitting is high, and the verification effect in the expanded queue is poor. Therefore, the combination of the plasma protein molecule and clinical data has excellent advantages in the prediction of chemotherapy sensitivity of pancreatic cancer patients, which far exceeds the current standard method used in clinic.

TABLE 1

Protein combinations	AUC
		GAI	0.810
GAF	0.778
		GAM	0.890
GIF	0.833
		GIM	0.865
GFM	0.888
		AIF	0.835
AFM	0.885
		IFM	0.903
AIM	0.808

Note: g: GSN, A: APOA4, I: IGHG1, M: IgM heavy chain, F: FCN 2.

In conclusion, the invention discovers for the first time that a chemotherapy sensitivity prediction model based on human plasma protein molecules (comprising GSN, APOA4, IGHG1, immunologlulin mu heavy chain and FCN2, any combination of the five proteins and any combination of parts of the five proteins) has a good prediction effect in pancreatic cancer patients, so that the chemotherapy sensitivity prediction model can be used as an important supplement of the existing chemotherapy response prediction strategy, and the efficiency is far superior to that of the existing clinical method.

Claims

1. A plasma protein molecule for detecting cancer chemotherapy sensitivity, which comprises at least one of human plasma proteins GSN, APOA4, IGHG1, Immunoglobulin mu heavy chain and FCN 2.

2. The plasma protein molecule of claim 1, comprising one of the following combinations of proteins: GAI, GAF, GAM, GIF, GIM, GFM, AIF, AFM, IFM, AIM, wherein G: GSN, A: APOA4, I: IGHG1, M: immunoglobulin mu heavy chain, F: FCN 2.

3. The plasma protein molecule of claim 2, comprising five proteins selected from the group consisting of human plasma proteins GSN, APOA4, IGHG1, Immunoglobulin mu heavy chain and FCN 2.

4. Use of a plasma protein molecule according to any one of claims 1 to 3 as a target for detecting sensitivity to cancer chemotherapy.

5. The use of claim 4, wherein the cancer species is pancreatic cancer.

6. The use of claim 5, wherein the chemotherapeutic regimen is a FOLFIRINOX regimen.

7. The use of claim 6, comprising measuring the concentrations of human plasma proteins GSN, APOA4, IGHG1, Immunoglobulin mu heavy chain and FCN2 in plasma, and inputting said 5 human plasma protein concentration data into a pre-generated random forest model for predictive assessment of the patient's chemotherapeutic response.

8. The application as claimed in claim 7, wherein the pre-generated random forest model is constructed by:

9. A kit for detecting cancer chemotherapy susceptibility, comprising:

(1) reagents for extracting proteins from plasma;