CN102183662A - Method for establishing colon cancer prognosis prediction model - Google Patents

Abstract

The invention provides a method for establishing a prognosis prediction model of colorectal cancer. The expression levels of SPARCL1 and P53 proteins in colorectal cancer tissues are detected by immunohistochemistry; the tissue expression levels of SPARCL1 and P53 proteins are graded by a semi-quantitative method; SPARCL1 and P53 The protein expression level was analyzed and verified by the combination of support vector machines, and finally a discriminant model was established. The invention combines immunohistochemical detection, marker combination and support vector machine analysis, and is jointly applied to establish a prediction model of colorectal cancer. Research shows that the present invention uses the combination of SPARCL1 and P53 as markers to construct a model, which has the function of experimentally assisting in predicting the prognosis of patients with colorectal cancer. It can be applied in the experiment of predicting the risk of metastasis and recurrence in patients with colorectal cancer after surgery.

Description

A method for establishing a prognosis prediction model of colorectal cancer

technical field

The invention belongs to the field of biotechnology, and in particular relates to a method for establishing a colorectal cancer prognosis prediction model using SPARCL1 and P53 as markers.

Background technique

Colorectal cancer is a common malignant tumor, and its morbidity and mortality rank among the top in the spectrum of malignant tumors in my country and western developed countries. From 2000 to 2004, the incidence rate of colorectal cancer ranked third among all malignant tumors in the United States, and the mortality rate also ranked third. In my country, with the improvement of living standards and changes in eating habits, the incidence of colorectal cancer is also increasing day by day, and has jumped to the fourth place. There are 400,000 new cases of colorectal cancer every year, with an increase rate of nearly 5%. Many of them It is a middle-aged person aged 30-40. Compared with the 1970s in the 1990s, the incidence of colorectal cancer in my country increased by 31.95% in urban areas and 8.51% in rural areas, and its mortality rate ranked fourth or fifth in the death spectrum of malignant tumors. It can be predicted that the incidence and mortality of colorectal cancer in my country will continue to rise steadily for a long period of time in the future, becoming one of the most common malignant tumors with the fastest rising incidence in my country.

In recent years, with the rapid development of science and technology, traditional treatment methods such as surgery, chemotherapy, and radiotherapy have continued to develop, and new drugs for biological targeted therapy have been continuously developed. However, the 5-year survival rate of colorectal cancer is still only about 63.4%. The speed obviously cannot keep up with the progress of treatment methods. At present, the method of judging the prognosis of colorectal cancer mainly relies on traditional staging methods, but many problems in clinical work cannot be fully explained by these traditional staging methods: why do patients with the same stage of colorectal cancer often have relatively different survival conditions? big difference? Why did the patients with stage I at the first diagnosis develop distant metastases later? Therefore, it is extremely important to find more markers and models related to the prognosis of colorectal cancer, to more accurately predict the prognosis of colorectal cancer patients and their sensitivity to different treatments, and to find targets for intervention and blocking to implement different treatments Strategies to achieve individualized treatment for patients.

P53 is the most important tumor suppressor gene discovered so far, located on human chromosome 17p13, containing 11 exons and 10 introns. It is divided into two types: wild type and mutant type. The wild type P53 gene not only plays a negative regulatory role as a tumor suppressor gene, but also participates in transcription, DNA damage and repair, cell cycle apoptosis control, cell apoptosis, cell proliferation, and cell differentiation. The process has the function of "molecular police". The mutant P53 protein is not easily hydrolyzed and has a longer half-life, and accumulates in malignant tumor cells. The mutant P53 protein can combine with the wild-type P53 protein to lose its negative regulation of cell growth. The overexpression of mutant P53 protein is closely related to the mutation of P53 gene. Previous studies have shown that P53 is related to the stage of patients, multidrug resistance, response to chemotherapy or radiotherapy, and postoperative recurrence and metastasis.

SPARCL1 belongs to the adhesion molecules that mediate cell-matrix interactions. SPARCL1 was first discovered in 1993 by Schraml et al. in the study of non-small cell lung cancer, named MAST9. The SPARCL1 protein belongs to the SPARC family and has 62% homology with the SPARC (secret protein acidic and rich in cysteine) sequence. Both have cysteine rich follistatin-like (FS) domain and extracellular calcium binding (EC) domain, but the N-terminus of SPARCL1 is much longer than SPARC, which is also due to the high structural homology with SPARC So named SPARC-like 1. The function of SPARCL1 has not been fully elucidated yet. SPARCL1 was downregulated in non-small cell lung cancer, metastatic prostate cancer, colorectal cancer, bladder cancer, and pancreatic ductal carcinoma, but upregulated in liver cancer. However, there is no clear report on the expression of SPARCL1 in colorectal cancer and its relationship with clinical features such as prognosis of colorectal cancer.

Bioinformatics is an emerging interdisciplinary subject that uses computers to store, retrieve and analyze biological information in life science research. The basic research process is: use computer to store nucleic acid or protein information, study scientific algorithms, compile corresponding software to analyze, compare and predict the information, and discover laws from it. Support vector machine (support vector machine, SVM) is a new classification technology proposed by Vapnik et al. in 1995, which has been widely used in many fields such as facial recognition and genomics.

Contents of the invention

The purpose of the present invention is to provide a method for establishing a prognosis prediction model of colorectal cancer. That is, it is a method to establish a marker combination model through the joint application of immunohistochemical detection, marker combination and support vector machine analysis. This method is implemented through the following steps:

1. Immunohistochemical method was used to detect the expression levels of SPARCL1 and P53 proteins in colorectal cancer tissues: tissue wax block specimen collection→slicing→overnight oven→dewaxing→hydration→antigen retrieval→natural cooling→washing→spin-drying, set In the wet box → drop A reagent (Endogenous Peroxidase Blocking Solution endogenous peroxidase blocker) to stand → wash the film → drop B reagent (Blocking Solution Non-Immune Serum non-immune animal serum) to stand → drop Primary antibody, standing→washing→dropping C reagent (Biotinylated Second Antibody Goat anti Mouse IgG biotinylated secondary antibody), standing→washing→dropping D reagent (Enzyme Conjugate HRP-Streptavidin Streptavidin protein-peroxidase), let stand→wash the film→DAB color development→wash after observing the color development→nucleus staining→wash→differentiate→wash with tap water→turn blue→dehydrate→dry→sealing;

2. The tissue expression level of SPARCL1 and P53 protein is graded by semi-quantitative method: 5 high-power fields of view are randomly detected for each sample, and the positive range and staining depth are scored respectively: the positive range is divided into 0-4 grades: 0 (<10% ), 1 (10%-25%), 2 (25%-50%), 3 (50%-75%), 4 (>75%); the staining depth is divided into 0-3 levels: 0 is negative, 1 It is light yellow, 2 is brownish yellow, and 3 is dark brown. Finally, the scores of the first two parts are added together, and the final result is divided into grades 0-7;

3. The expression levels of SPARCL1 and P53 proteins were analyzed and verified by the support vector machine combination, and finally the discriminant model was established: the support vector machine adopts the radial basis kernel function (radial based kernel), the Gamma value is set to 0.6, and the penalty function (C) is set to for 19. The selection of feature vectors adopts the method of statistical filtering combined with model-dependent screening. The combination of SPARCL1 and P53 was used as the input of the support vector machine model, and the prediction effect of the model (three-year survival/death) was evaluated by the ten-fold cross-validation method, and the Youden index (Youden index = [(Sensitivity + Specificity)/2+Sensitivity]/2) The highest combination is used as the final model, and the predicted value after ten-fold cross-validation is used as the final result.

Another object of the present invention is to provide the application of the model in the risk prediction experiment of postoperative metastasis and recurrence of colorectal cancer patients.

In the present invention, the combination of SPARCL1 and P53 is used as a marker to construct a model, which is used in an experiment for predicting the risk of metastasis and recurrence in patients with colorectal cancer after surgery. The invention combines immunohistochemical detection, marker combination and support vector machine analysis, and is jointly applied to establish a prediction model of colorectal cancer. After research, the expression of SPARCL1 and P53 in the resected tumor tissue of colorectal cancer patients was detected by immunohistochemistry. For the first time, the expression values of the two were combined by bioinformatics to form a SPARCL1/P53 combination model, and the prognostic model was confirmed. It has the role of experimental assistant in predicting the prognosis of patients with colorectal cancer. Thus, for the first time, a method for establishing a marker combination model was discovered through the joint application of immunohistochemical detection, marker combination and support vector machine analysis.

Description of drawings

Figure 1 shows the survival curves of patients (all stages) with different discrimination results of the SPARCL1/P53 prognostic model.

Figure 2 shows the survival curves of stage II and stage III patients.

Figure 3 shows the survival curves of patients (stage II+III) with different discrimination results of the SPARCL1/P53 prognostic model.

Figure 4 shows the survival curves of patients (stage II) with different discrimination results of the SPARCL1/P53 prognostic model.

Figure 5 shows the survival curves of patients (stage III) with different discrimination results of the SPARCL1/P53 prognostic model.

Detailed ways

The present invention will be further elaborated below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these specific implementations are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

Example 1

Step 1: Detection of SPARCL1 and P53 proteins in colorectal cancer tissues

Immunohistochemical methods were used to detect the expression of P53 and SPARCL1 in the surgically resected colorectal cancer tissues of patients.

期43例，

期56例，

期9例），且诊断均经术后病理证实，所有患者的临床病理资料术后均以随访表形式登记，并每年通过信件或电话随访其生存及复发转移等情况 36 个月以上，本研究实验开始前再次随访核实。The 131 colorectal cancer tissue wax blocks were selected from surgically resected tissues of colorectal cancer patients admitted to the Second Affiliated Hospital of Zhejiang University School of Medicine from 1999 to 2004 (23 cases of stage I,

Period 43 cases,

Period 56 cases,

9 cases in the first stage), and the diagnosis was confirmed by postoperative pathology. The clinicopathological data of all patients were registered in the form of follow-up form after operation, and their survival, recurrence and metastasis were followed up for more than 36 months by letter or telephone every year. Check again before the experiment started.

All tissue wax block specimens were from the tissue wax blocks archived in the Pathology Department of the Second Affiliated Hospital of Zhejiang University School of Medicine. After the tissue was removed during the operation, it was immediately fixed in 4% formalin. After sufficient fixation, the tissue was collected, dehydrated, transparent, soaked in wax, and embedded. The section was confirmed pathologically by HE staining, and then stored for a long time.

Immunohistochemical kits include ready-to-use fast immunohistochemical MaxVision kit (mouse/rabbit, Maixin Bio, catalog number: KIT-5010), immunohistochemical SP hypersensitivity kit (sheep, Maixin Bio, catalog number : KIT-9709) and DAB Chromogenic Kit (Maixin Bio, catalog number: DAB-0031).

Antibodies used include: P53 antibody (mouse anti-human P53 monoclonal antibody, Zhongshan Jinqiao, catalog number: ZM-0408, ready-to-use), SPARCL1 antibody (Polyclonal goat anti-human SPARC-like 1 antibody (R&D, catalog number: AF2728, dilution 1:160).

Immunohistochemical process: Cut the tissue wax block into about 4 μm thick slices, spread the slices and mount them on the glass slides pretreated with polylysine → oven at 59°C overnight → xylene dewaxing (20min×3 → gradient alcohol water Hydration (absolute wine 5min×3→95% alcohol 5min→75% alcohol 5min)→hydration in distilled water→1×EDTA, heating at 98°C for 15 minutes (antigen retrieval)→natural cooling to room temperature→washing in distilled water→ Wash the film with TBS solution (PH 7.4) (5min) → shake dry, place in a wet box → add reagent A (endogenous peroxidase blocker) dropwise, let stand for 10min → wash the film with TBS for 3min×3 times → drop Add reagent B (non-immune animal serum) and let it stand for 15 minutes→drain excess liquid, add primary antibody dropwise, let stand for 2 hours→TBS 5min×3 times→add reagent C (biotin-labeled secondary antibody), let stand for 15min → TBS 5min × 2 times → drop D reagent (streptomyces avidin-peroxidase), let stand for 15min → TBS 3min × 3 times → drop freshly prepared DAB color reagent → about 1min, observe After coloring, place in tap water and rinse→soak in hematoxylin for 10min→tap water rinse→hydrochloric acid alcohol differentiation for 1s→tap water rinse→soak in 56℃ hot water to turn blue→gradient alcohol dehydration (75% alcohol for 5min→95% alcohol for 5min→none Water alcohol 5min×3)→Blow dry, and seal with neutral resin.

Step 2: SPARCL1, P53 protein expression level grading

In this embodiment, the results of step 1 are graded.

Five high-power fields of view (×400 times) were randomly detected for each specimen, and the semi-quantitative method was used to comprehensively consider the range of positive cells and the depth of staining. The positive range is divided into 0-4 levels: 0 (<10%), 1 (10%-25%), 2 (25%-50%), 3 (50%-75%), 4 (>75%); The staining depth is divided into 0-3 grades: 0 is negative, 1 is light yellow, 2 is brownish yellow, and 3 is dark brown. Add the scores of the first two parts, and the final result is scored on a scale of 0-7.

Step 3: The expression levels of SPARCL1 and P53 proteins were combined to form a model by bioinformatics

The experimental data were analyzed by the ZUCI-ProteinChip Data Analyze System software package designed by Yu Jiekai, Zhejiang University Cancer Institute. The discriminant model was established by the support vector machine method, and the ten-fold cross-validation method was used as a method to evaluate the discriminant effect of the model. The support vector machine adopts the radial basis kernel function (radial based kernel), the Gamma value is set to 0.6, and the penalty function (C) is set to 19. The selection of feature vectors adopts the method of statistical filtering combined with model-dependent screening. The combination of SPARCL1 and P53 was used as the input of the support vector machine model, and the prediction effect of the model (three-year survival/death) was evaluated by the ten-fold cross-validation method, and the Youden index (Youden index = [(Sensitivity + Specificity)/2+Sensitivity]/2) The highest combination is used as the final model, and the predicted value after ten-fold cross-validation is used as the final result. According to the SPARCL1/P53 prediction model, patients can be divided into two groups, namely good prognosis (predicted value = 0) and bad prognosis (predicted value = 1).

Example 2

1. The role of SPARCL1/P53 model in judging the prognosis of colorectal cancer patients (all stages)

Figure 1 shows the survival curves of patients (all stages) with different discrimination results of the SPARCL1/P53 prognostic model. The Kaplan-Meier method validation showed that the SPARCL1/P53 prediction model could divide 131 patients into two groups with great differences in survival time (P<0.001), namely good prognosis (predicted value = 0) and bad prognosis (predicted value = 0). 1), the estimated median survival time of patients in the good prognosis group was 91.676 months, and the estimated median survival time of patients in the bad prognosis group was 41.928 months.

． Predictive effect of SPARCL1/P53 model on postoperative recurrence and metastasis of colorectal cancer patients

Among these patients, the incidence of postoperative recurrence and metastasis in the good prognosis group (prediction value = 0) was 22.34%, while that in the bad prognosis group (prediction value = 1) was 64.86%, P<0.001 for the comparison between the two .

Predictive effect of prognosis model on postoperative recurrence and metastasis

have recurrence and metastasis No recurrence and metastasis Prediction(0) 73 twenty one Prediction(1) 13 twenty four

Prediction: the predictive value of each marker combination model for prognosis

(0=good prognosis, 1=bad prognosis).

． The role of SPARCL1/P53 model in judging the prognosis of patients with stage II and stage III colorectal cancer

A total of 99 cases of stage II and stage III colorectal cancer were selected from a total of 131 cases. Combined with traditional staging, the prediction effect of the SPARCL1/P53 model on the survival of patients with stage II and stage III colorectal cancer was observed.

Figure 2 shows the survival curves of stage II and stage III patients. Kaplan-Meier survival analysis showed that the estimated median survival time of patients with stage Ⅱ and stage Ⅲ was not significantly different, 80.566 months VS 63.038 months (P=0.039).

Figure 3 shows the survival curves of patients (stage II+III) with different discrimination results of the SPARCL1/P53 prognostic model. The SPARCL1/P53 model divided 99 patients with stage II and stage III patients into two groups with large differences in survival time (P<0.001), namely good prognosis (predicted value = 0) and bad prognosis (predicted value = 1), where good The estimated median survival time of the patients in the prognosis group was 81.658 months, while the estimated median survival time of the patients in the bad prognosis group was 43.889 months.

Figure 4 shows the survival curves of patients (stage II) with different discrimination results of the SPARCL1/P53 prognostic model. In 43 patients with stage II colorectal cancer, the SPARCL1/P53 model can also divide the patients into two groups with a large difference in survival time (P<0.001), which are good prognosis (predicted value = 0) and bad prognosis (predicted value =1), the estimated median survival time of patients in the good prognosis group was 88.65 months, and the estimated median survival time of patients in the bad prognosis group was 50.509 months.

Figure 5 shows the survival curves of patients (stage III) with different discrimination results of the SPARCL1/P53 prognostic model. In 56 patients with stage III colorectal cancer, the SPARCL1/P53 prognostic model can also divide the patients into two groups with a large difference in survival time (P<0.001), which are good prognosis (predicted value = 0) and bad prognosis (predicted value = 0). value = 1), the estimated median survival time of patients in the good prognosis group was 75.74 months, and the estimated median survival time of patients in the bad prognosis group was 36.167 months.

Claims (2)

1. A method for establishing a colorectal cancer prognosis prediction model, characterized in that, it is realized by the following steps: (1) Immunohistochemical method was used to detect the expression levels of SPARCL1 and P53 proteins in colorectal cancer tissues: collect tissue wax block specimens → section → oven overnight → dewaxing → hydration → antigen retrieval → natural cooling → washing → drying, Put it in a wet box → add endogenous peroxidase blocking agent dropwise, let it stand → wash the film → add non-immune animal serum, let it stand → add primary antibody, let it stand → wash the film → add biotin-labeled Secondary antibody, stand still → wash the film → drop streptavidin-peroxidase, stand still → wash the film → DAB color development → observe the color development and then rinse → nuclear staining → rinse → differentiation → tap water rinse →Return to blue→Dehydration→Blow dry→Seal; (2) The tissue expression levels of SPARCL1 and P53 proteins were graded by semi-quantitative method: 5 high-power fields of view were randomly detected for each sample, and the positive range and staining depth were scored respectively: the positive range was divided into 0-4 grades: <10% was Grade 0, 10%-25% is grade 1, 25%-50% is grade 2, 50%-75% is grade 3, >75% is grade 4; staining depth is divided into 0-3 grades: 0 is negative, 1 It is light yellow, 2 is brownish yellow, and 3 is dark brown; finally, the scores of the first two parts are added together, and the final result is divided into 0-7 grades; (3) The expression levels of SPARCL1 and P53 proteins were analyzed and verified by the support vector machine combination, and a discriminant model was established: the support vector machine used the radial basis kernel function, the Gamma value was set to 0.6, the penalty function (C) was set to 19, and the feature vector The selection adopts the method of statistical filtering combined with model-dependent screening. The combination of SPARCL1 and P53 is used as the input of the support vector machine model, and the prediction effect of the model is evaluated by the ten-fold cross-validation method. Three-year survival/death is selected to establish a support vector The Youden index predicted by the machine model: Youden index=[(sensitivity+specificity)/2+sensitivity]/2), the highest combination is used as the final model, and the predicted value after ten-fold cross-validation is used as the final result. 2. The application of the model established by the method according to claim 1 in the risk prediction experiment of postoperative metastasis and recurrence of patients with colorectal cancer.

Images