Prediction model for lung squamous cell carcinoma

Abstract

The invention relates to a prediction model for lung squamous cell carcinoma, and belongs to the technical field of disease prediction. On the basis of proteomics, protein molecule changes with specific execution functions in the lung squamous cell carcinoma occurrence process are screened, and a lung squamous cell carcinoma occurrence prediction model is constructed in combination with multiple machine learning models and public data. The prediction model for lung squamous cell carcinoma occurrence comprises six prediction factors: DDAH1, ATP11A, NME3, PON2, SLC1A5 and LPCAT1, and the prediction effectiveness (AUC value) of the prediction model can reach 97.6% and is higher than the prediction AUC value of a single factor.

Description

Technical Field

[0001] The present invention relates to the technical field of disease prediction, and in particular to a prediction model for the occurrence of lung squamous cell carcinoma. Background Art

[0002] The latest data show that lung cancer is still the malignant tumor with the highest incidence and mortality in my country. The pathological classification of lung cancer mainly includes major pathological types such as non-small cell lung cancer, small cell carcinoma and large cell carcinoma. Among them, non-small cell lung cancer accounts for about 80-85% of lung cancer. Squamous cell carcinoma of the lung, as the main histological type of non-small cell lung cancer, accounts for about 20%-30%, and the 5-year survival rate is less than 20%. Compared with lung adenocarcinoma, squamous cell carcinoma has poor treatment effect and few treatment options. Due to the low positive rate of driver gene mutation in squamous cell carcinoma of the lung, patients are often older, at a later stage, and have more complications at diagnosis. Most tumors are located in the center, and most tumors are in the middle and late stages, which makes treatment more difficult. The median survival of patients is about 30% shorter than that of patients with other NSCLC subtypes. Therefore, screening and early diagnosis and treatment are effective measures to reduce the mortality rate of lung cancer in the population. Although the genetic changes in the occurrence and development of squamous cell carcinoma of the lung have been revealed, according to clinical data, the incidence of squamous cell carcinoma of the lung is still high, and accurate prediction of the occurrence of squamous cell carcinoma of the lung is still a problem that needs to be solved urgently in clinical practice.

[0003] Although there are many prediction models for the occurrence of squamous cell lung carcinoma, most of them are single gene or transcription level predictions, which cannot predict the changes in specific functional molecules in the tumor. Squamous cell lung carcinoma is still one of the main pathological types of lung cancer, so a more accurate prediction model is needed. Summary of the invention

[0004] The purpose of the present invention is to provide a prediction model for the occurrence of lung squamous cell carcinoma based on proteomics, which has high prediction efficiency (AUC value) and accurate prediction results.

[0005] In order to achieve the above object, the present invention provides the following technical solutions: A prediction model for the occurrence of lung squamous cell carcinoma, wherein the construction process of the prediction model comprises the following steps: S1. Sample data collection: Collect lung cancer related proteomic data and metabolism related genes, wherein the lung cancer related proteomic data include lung adenocarcinoma tissue protein data, lung adenocarcinoma paracancerous tissue protein data, lung squamous cell carcinoma tissue protein data, and lung squamous cell carcinoma paracancerous tissue protein data, and the data sources of the lung cancer related proteomic data include self-test data and public data; S2. Screening of metabolic genes specifically expressed in squamous cell lung carcinoma; for the proteomics data collected, use the rank sum test method to screen the protein molecules differentially expressed in squamous cell lung carcinoma tissue and lung adenocarcinoma tissue compared with adjacent tissue; based on the obtained differential proteins, screen the protein molecules differentially expressed only in squamous cell lung carcinoma tissue compared with lung adenocarcinoma tissue; based on the collected metabolism-related genes, screen the differentially expressed metabolism-related protein molecules specific to squamous cell lung carcinoma; take the intersection of the differentially expressed metabolism-related protein molecules from the self-test data source and the public data source, and screen the differentially expressed protein molecules specific to squamous cell lung carcinoma; S3. Use machine learning models to publicly screen candidate predictive markers and build a predictive model. Based on the differentially expressed protein molecules specifically expressed in lung squamous cell carcinoma screened in S2, use support vector machine, lasso regression, random forest and Boruta algorithms to screen candidate predictive markers in the protein data of self-tested lung squamous cell carcinoma and adjacent tissues, and screen out candidate markers that are significantly underexpressed or significantly overexpressed in lung squamous cell carcinoma tissues. Based on the self-test data, use lasso regression again to screen candidate markers, determine the final predictive markers, and use the logistic regression model to build a lung squamous cell carcinoma occurrence prediction model based on the self-test data.

[0006] Furthermore, the process of constructing the prediction model also includes: S4. Validation of the prediction model effectiveness by an independent cohort: Based on the final prediction marker, the proteomic data of lung squamous cell carcinoma from a public source were used as an independent data set to validate the effectiveness of the model. Based on the protein expression level of the final prediction marker, the independent data set was randomly divided into a training group and a test group at a ratio of 7:3 using random forest to construct a prediction model, and the ROC curve was plotted to calculate the AUC value, and the difference between multi-gene joint diagnosis and single-gene diagnosis was compared.

[0007] Furthermore, there are 4875 metabolism-related genes, and the sources of the metabolism-related genes are Human-GEM, KEGG, Reactome, Cancer cell Metabolism Gene and Metabolic enzyme and transport.

[0008] Further, in step S2, for the self-tested proteomics data, the conditions for screening the protein molecules differentially expressed in lung squamous cell carcinoma tissue and lung adenocarcinoma tissue compared with adjacent cancer tissue are: |log2(FC)|>1, P<0.05; For publicly collected proteomics data, the conditions for screening protein molecules that are differentially expressed in lung squamous cell carcinoma tissues and lung adenocarcinoma tissues compared with adjacent cancer tissues are: |log2(FC)|>2, FDR<0.01.

[0009] The beneficial effects of the present invention are: Based on proteomics, the present invention screened the changes in protein molecules that perform specific functions in the occurrence of lung squamous cell carcinoma, and combined multiple machine learning models and public data to construct a prediction model for the occurrence of lung squamous cell carcinoma. The prediction model for the occurrence of lung squamous cell carcinoma in the present invention includes 6 prediction factors: DDAH1, ATP11A, NME3, PON2, SLC1A5, LPCAT1, and its prediction efficiency (AUC value) can reach 97.6%, which is higher than the prediction AUC value of a single factor.

[0010] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, some of the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0012] Figure 1 A flowchart constructed for the prediction model shown in the present invention; Figure 2 A schematic diagram of protein expression values ​​in a preferred embodiment of the present invention; Figure 3 The ROC curve of the self-test data test set and the external validation data in a preferred embodiment of the present invention; Figure 4 is the expression value of the predicted gene in the independent validation set in a preferred embodiment of the present invention; Figure 5 A Gini coefficient (importance) diagram of predicted genes in an independent validation set in a preferred embodiment of the present invention; Figure 6 This is a ROC curve diagram of the joint prediction performance in an independent validation set in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0013] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0014] The technical solutions described in the present invention, unless otherwise specified, are all conventional solutions in the art; the reagents or materials described, unless otherwise specified, are all from commercial channels.

[0015] Example 1 Please refer to Figure 1 The present invention provides a prediction model for the occurrence of lung squamous cell carcinoma, and the construction process of the prediction model includes the following steps: S1, Sample and data collection; 1) Protein data: This paper collected the proteomic database of 120 lung cancer patients that had been tested in the early stage, including 30 cases of lung adenocarcinoma tissue (LUAD) and paracancerous tissue, 30 cases of lung squamous cell carcinoma tissue (LUSC) and paracancerous tissue, and then collected the LUSC protein data from two published public sources, including 216 cases of LUSC and 99 cases of paracancerous tissue, and the LUAD protein data from CPTAC data sources, including 110 cases of LUAD and 102 cases of paracancerous tissue; 2) Metabolism-related genes: A total of 4,875 metabolism-related genes were collected from Human-GEM, KEGG, Reactome, Cancer cell Metabolism Gene databases and publicly published Metabolic enzyme and transport for downstream analysis and metabolism-related genes.

[0016] S2. Screening of metabolic genes specifically expressed in LUSC: 1) For the self-tested proteomics data, the rank sum test method was used to screen the differentially expressed protein molecules in LUSC and LUAD compared with adjacent tissues (|log2(FC)|>1, P<0.05); based on the obtained differential proteins, the protein molecules that were differentially expressed only in LUSC compared with LUAD were screened; finally, based on the collected 4875 metabolism-related genes, the differentially expressed metabolism-related protein molecules specific to LUSC were screened; 2) For the publicly collected LUSC and LUAD data, the same differential molecules were used to screen the differential proteins between LUAS and LUAD compared with adjacent adjacent tumors. The screening conditions were: |log2(FC)|>2, FDR<0.01; the rest of the processing methods were the same; 3) Finally, the intersection of the differential metabolism-related protein molecules from the two data sources was taken, and 34 LUSC-specific differentially expressed protein molecules were finally screened.

[0017] S3. Use multiple machine learning models to publicly screen candidate predictive markers and build predictive models: 1) Please refer to Figure 2-Figure 3Based on the above-screened 34 LUSC-specific differentially expressed protein molecules, the candidate predictive markers were screened using support vector machine, lasso regression, random forest, and Boruta algorithms in the protein data of 60 self-tested LUSC and adjacent cancer tissues. Finally, a total of 7 candidate markers were screened, namely CA12, DDAH1, ATP11A, NME3, PON2, SLC1A5, and LPCAT1; among these 7 protein molecules, ATP11A, LPCAT1, DOAH1, PON2, and NME3 were significantly underexpressed in LUSC, while SLC1A5 and CA12 were significantly overexpressed in LUSC.

[0018] 2) Based on the self-test data, lasso regression was used again to screen candidate markers, and DDAH1, ATP11A, NME3, PON2, SLC1A5, and LPCAT1 were determined as the final predictive markers. The logistic regression model was used to construct a LUSC occurrence prediction model based on self-test data. The publicly published LUSC data was used as an external validation set, and the ROC curve was drawn. The results showed that the prediction performance (AUC value) of these 6 candidate markers based on self-test data can reach 97.9%, and the AUC value of the external validation data is 97.58%.

[0019] S4. Independent cohort validation of the prediction model: 1) Please refer to Figure 4-Figure 6 Based on the six markers for predicting LUSC selected from the self-test data, the model was validated using publicly published proteomic data of LUSC (108 cases of LUSC and 99 cases of paracancerous tissues) as an independent data set. First, the expression of these six genes in the data set was detected. It was found that compared with paracancerous tissues, these six genes ATP11A, LPCAT1, DOAH1, PON2 and NME3 were significantly underexpressed in LUSC, while SLC1A5 was significantly overexpressed in LUSC, which was consistent with the results of the self-test data. 2) Based on the protein expression levels of these six candidate markers, the data were randomly divided into a training group and a test group at a ratio of 7:3 using random forest to build a prediction model, draw the ROC curve, calculate the AUC value, and compare the difference between multi-gene joint diagnosis and single gene diagnosis. The results showed that ATP11A, LPCAT1, DOAH1, and SLC1A5 were more important in the independent test set, but the joint prediction performance (AUC value) of the six markers could reach 98.6%.

[0020] S5. The present invention establishes a prediction model for the occurrence of LUSC. Based on the verification of the above self-test data and independent public data, a prediction model for the occurrence of LUSC based on proteomics was established, including 6 prediction factors: DDAH1, ATP11A, NME3, PON2, SLC1A5, and LPCAT1. Its prediction efficiency (AUC value) can reach 97.6%.

[0021] In summary, the present invention screened the changes in protein molecules that perform specific functions in the occurrence of lung squamous cell carcinoma based on proteomics, and constructed a prediction model for the occurrence of lung squamous cell carcinoma by combining multiple machine learning models and public data. The prediction model for the occurrence of lung squamous cell carcinoma in the present invention includes 6 prediction factors: DDAH1, ATP11A, NME3, PON2, SLC1A5, LPCAT1, and its prediction efficiency (AUC value) can reach 97.6%, which is higher than the prediction AUC value of a single factor.

[0022] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0023] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A prediction model for the occurrence of lung squamous cell carcinoma, characterized in that: The construction process of the prediction model includes the following steps: S1. Sample data collection: Collect lung cancer related proteomic data and metabolism related genes, wherein the lung cancer related proteomic data include lung adenocarcinoma tissue protein data, lung adenocarcinoma paracancerous tissue protein data, lung squamous cell carcinoma tissue protein data, and lung squamous cell carcinoma paracancerous tissue protein data, and the data sources of the lung cancer related proteomic data include self-test data and public data; S2. Screening of metabolic genes specifically expressed in lung squamous cell carcinoma. For the collected proteomics data, the rank sum test method was used to screen the protein molecules that were differentially expressed in lung squamous cell carcinoma tissues and lung adenocarcinoma tissues compared with adjacent tissues. Based on the obtained differential proteins, screen the protein molecules that are differentially expressed only in lung squamous cell carcinoma tissue compared with lung adenocarcinoma tissue; based on the collected metabolism-related genes, screen the differentially expressed metabolism-related protein molecules specific to lung squamous cell carcinoma; take the intersection of the differentially expressed metabolism-related protein molecules from the self-test data source and the public data source, and screen the differentially expressed protein molecules specific to lung squamous cell carcinoma; S3. Use machine learning models to publicly screen candidate predictive markers and build a predictive model: Based on the differentially expressed protein molecules specifically in squamous cell lung cancer screened in S2, use support vector machine, lasso regression, random forest and Boruta algorithms to screen candidate predictive markers in the protein data of self-tested squamous cell lung cancer and adjacent tissues, and screen out candidate markers that are significantly under-expressed or significantly over-expressed in squamous cell lung cancer tissues; Based on the self-test data, lasso regression was used again to screen candidate markers, determine the final predictive markers, and use the logistic regression model to construct a prediction model for lung squamous cell carcinoma based on self-test data.

2. The prediction model for the occurrence of lung squamous cell carcinoma according to claim 1, characterized in that: The process of constructing the prediction model also includes: S4. Validation of the prediction model effectiveness by an independent cohort: Based on the final prediction marker, the proteomic data of lung squamous cell carcinoma from a public source were used as an independent data set to validate the effectiveness of the model. Based on the protein expression level of the final prediction marker, the independent data set was randomly divided into a training group and a test group at a ratio of 7:3 using random forest to construct a prediction model, and the ROC curve was plotted to calculate the AUC value, and the difference between multi-gene joint diagnosis and single-gene diagnosis was compared.

3. The prediction model for the occurrence of lung squamous cell carcinoma according to claim 1, characterized in that: There are 4875 metabolism-related genes, and the sources of the metabolism-related genes are Human-GEM, KEGG, Reactome, Cancer cell MetabolismGene and Metabolic enzyme and transport.

4. The prediction model for the occurrence of lung squamous cell carcinoma according to claim 1, characterized in that: In step S2, for the self-tested proteomics data, the conditions for screening the protein molecules differentially expressed in lung squamous cell carcinoma tissue and lung adenocarcinoma tissue compared with adjacent tissues are: |log2(FC)| > 1, P < 0.05; For publicly collected proteomic data, the conditions for screening protein molecules that are differentially expressed in lung squamous cell carcinoma tissues and lung adenocarcinoma tissues compared with adjacent adjacent tissues are: |log2(FC)| > 2, FDR < 0.01.

Images