Cell membrane protein pump efficiency monitoring system for solid tumor brain metastasis risk assessment

Abstract

The invention relates to the technical field of solid tumor clinical diagnosis and treatment, and discloses a cell membrane protein pump efficiency monitoring system for solid tumor brain metastasis risk assessment. Comprising a biological sample processing module used for obtaining solid tumor biological sample cells, preparing biological sample cells to be detected and obtaining positive control sample cells; the image acquisition module is used for performing protein detection on the to-be-detected biological sample cells and the positive control sample cells to obtain a detection image; the fractional enhanced image learning module is used for carrying out third-order fractional enhanced image learning on the detection image, and the third-order fractional enhanced image learning comprises the following steps: carrying out secondary quantitative analysis on P-glycoprotein of a biological sample cell to be detected on a predicted 141 thousand Dalton site in the detection image; performing third feature image learning on the detection image in a fuzzy towing region appearing at a 150-300 thousand Dalton position on the membrane in the detection image; and the risk assessment module is used for comprehensively assessing the brain metastasis risk of the solid tumor based on the result of the third-order fractional enhanced image learning.

Description

Technical Field

[0001] This application relates to the field of clinical diagnosis and treatment technology for solid tumors, and in particular to a cell membrane protein pump efficiency monitoring system for assessing the risk of brain metastasis in solid tumors. Background Technology

[0002] Brain metastasis of solid tumors is a significant prognostic factor for patients with advanced cancer. In particular, EGFR-mutant tumors such as non-small cell lung cancer (NSCLC) often result in insufficient drug concentrations penetrating the blood-brain barrier due to the drug efflux effect mediated by P-glycoprotein (P-gp), exacerbating drug resistance in brain metastases. As an ATP-dependent efflux pump, P-gp's glycosylation modification leads to changes in protein molecular weight, causing a shift in the protein band position in traditional Western blotting (WB) experiments. Furthermore, the expression of P-gp varies among different tissues and cells, easily leading to false-negative results due to weak or absent expression.

[0003] Current technologies for P-gp detection largely rely on qualitative analysis of single molecular weight sites, failing to consider variations in molecular weight range due to glycosylation modifications (e.g., 160-200 kDa glycosylation) and tissue-specific expression differences, making it difficult to accurately identify weakly expressed samples. Furthermore, traditional detection methods do not employ multi-fractionation sample preparation or segmented image learning strategies, failing to address the interference of glycosylation modifications and fragmentation methods on detection results, resulting in a lack of quantitative indicators for brain metastasis risk assessment. Although current technologies recognize the association between P-gp and brain metastasis, a systematic detection scheme targeting glycosylation modifications and weak expression characteristics has not yet been proposed, particularly lacking a third-order reinforcement learning method that incorporates multi-dimensional image analysis.

[0004] Therefore, there is an urgent need to design a technical solution to solve at least one of the above-mentioned technical problems. Summary of the Invention

[0005] This application provides a cell membrane protein pump efficacy monitoring system for assessing the risk of brain metastasis in solid tumors. It aims to overcome the interference of glycosylation and provide a P-gp efficacy monitoring system that covers multiple expression states, thereby providing accurate quantitative assessment of the risk of brain metastasis in solid tumors.

[0006] In a first aspect, this application provides a cell membrane protein pump efficiency monitoring system for assessing the risk of brain metastasis in solid tumors, comprising:

[0007] The biological sample processing module is used to obtain solid tumor biological sample cells, prepare test biological sample cells using different lysis methods, and obtain positive control sample cells.

[0008] The image acquisition module is used to perform protein detection on the biological sample cells to be tested and the positive control sample cells, and to acquire detection images containing protein bands, wherein the protein detection includes the detection of P-glycoprotein;

[0009] The phased enhancement image learning module is used to perform three-stage phased enhancement image learning on the detection image, including: qualitative analysis of P-glycoprotein in the target biological sample cells at the predicted 141 kDalton site in the detection image, combined with the detection image of positive control sample cells; secondary quantitative analysis of P-glycoprotein in the target biological sample cells in the 160 to 200 kDalton range in the detection image, combined with the protein glycosylation characteristics and the detection images of the target biological sample cells prepared by different lysis methods, compared with the detection images of positive control sample cells; and third feature image learning on the detection image in the blurred drag region at the 150 to 300 kDalton position on the membrane, combined with the expression differences of P-glycoprotein in different tissue cells, to identify the weak expression or no expression of P-glycoprotein in the target biological sample cells.

[0010] The risk assessment module is used to comprehensively assess the risk of brain metastasis of solid tumors based on the results of the three-level enhanced image learning.

[0011] In some embodiments, obtaining solid tumor biological sample cells and preparing test biological sample cells using different lysis methods includes: isolating and obtaining biological sample cells from solid tumor patient samples, the samples including but not limited to cells derived from non-small cell lung cancer, liver cancer, and gastrointestinal tumors; selecting at least two different cell lysis methods to prepare test biological sample cells based on the characteristics of different tissue cells, the lysis methods including at least two of physical lysis, chemical lysis, or enzymatic lysis; and performing protein extraction and purification on the test biological sample cells prepared by different lysis methods to obtain a sample solution for detection.

[0012] In some embodiments, the step of detecting proteins in the test biological sample cells and positive control sample cells to obtain a detection image containing protein bands includes: separating the sample solutions of the test biological sample cells and the positive control sample cells by protein electrophoresis, allowing proteins to migrate in the gel according to their molecular weight; transferring the proteins in the gel to a solid membrane to form a protein membrane; then blocking the protein membrane, adding a specific antibody against P-glycoprotein for incubation, allowing the antibody to specifically bind to P-glycoprotein; and finally detecting the bound antibody by a colorimetric or fluorescent labeling method to obtain a detection image containing P-glycoprotein bands, wherein the detection image includes the distribution of protein bands corresponding to different lanes.

[0013] In some embodiments, the qualitative analysis of P-glycoprotein in the test biological sample cells at the predicted 141 kDalton site in the detection image, combined with the detection image of positive control sample cells, includes: locating a preset site region with a molecular weight of 141 kDalton in the detection image; extracting the band signal of the test biological sample cell lane within this site region and comparing it with the band signal of the positive control sample cell lane at the same site; if the test sample lane has a band at this site that matches the characteristics of the positive control signal, it is determined that the test sample contains the basic expression form of P-glycoprotein, thus completing the qualitative analysis.

[0014] In some embodiments, the secondary quantitative analysis of P-glycoprotein in the test biological sample cells, within the 160 to 200 kDalton range of the detection image, is performed by comparing the detection images of the test biological sample cells with those of the positive control sample cells, taking into account the protein glycosylation characteristics and different lysis methods. This includes: identifying band regions within the 160 to 200 kDalton molecular weight range in the detection image, since P-glycoprotein glycosylation may lead to an increase in molecular weight, the bands in this range reflect the protein form after glycosylation modification; extracting the band signal intensity and distribution characteristics of the test biological sample cells prepared by different lysis methods within this range, and comparing them with the band signals of the positive control sample cells in the same range using multiple lanes; determining the relative expression level of glycosylated P-glycoprotein in the test sample by calculating the ratio of the band signal intensity of the test sample to the signal intensity of the positive control, combined with the molecular weight shift law of glycosylation modification, thus completing the secondary quantitative analysis.

[0015] In some embodiments, the blurred banding region appearing at a position of 150 to 300 kilodaltons on the membrane in the detection image, combined with the expression differences of P-glycoprotein in different tissue cells, is used to perform a third feature image learning on the detection image to identify the weak or no expression of P-glycoprotein in the cells of the biological sample to be tested. This includes: performing image enhancement processing on the blurred banding region at a position of 150 to 300 kilodaltons in the detection image, including noise filtering, contrast adjustment, and edge detection; establishing a database of P-glycoprotein expression features of cells from different tissue sources, containing the banding morphology features of known weakly expressed or non-expressed samples; matching and comparing the processed features of the banding region of the sample to be tested with the features in the database; if a weak band or a banding signal of a specific morphology below a preset signal threshold is detected, and the signal appears in samples prepared by at least two lysis methods, it is determined to be weakly expressed P-glycoprotein; if no signal matching the features is detected, it is determined to be non-expressed, thus avoiding false negative results.

[0016] In some embodiments, the comprehensive assessment of the risk of brain metastasis of solid tumors based on the results of the three-stage enhanced image learning includes: integrating the presence of basic P-glycoprotein expression obtained from the first-stage qualitative analysis, the relative expression level of glycosylated proteins obtained from the second-stage quantitative analysis, and the identification results of weak expression or no expression obtained from the third-stage feature learning; weighting the analysis results of different stages according to a preset risk assessment model, wherein the expression level of glycosylated proteins and the identification results of weak expression are used as key risk factors; generating a risk level of brain metastasis of solid tumors based on the calculation results, wherein the risk level is correlated with the total expression level of P-glycoprotein, the degree of glycosylation modification, and tissue-specific expression differences, providing a quantitative basis for clinical assessment.

[0017] Secondly, this application provides a method for monitoring the efficiency of cell membrane protein pumps in assessing the risk of brain metastasis in solid tumors, applicable to the cell membrane protein pump efficiency monitoring system for assessing the risk of brain metastasis in solid tumors provided in any embodiment of this application. The method includes:

[0018] Solid tumor biological sample cells were obtained, and test biological sample cells were prepared using different lysis methods. Positive control sample cells were also obtained.

[0019] Protein detection is performed on the biological sample cells to be tested and the positive control sample cells to obtain detection images containing protein bands. The protein detection includes the detection of P-glycoprotein.

[0020] The detection images undergo three-stage enhanced image learning, including: qualitative analysis of P-glycoprotein in the target biological sample cells at the predicted 141 kDalton site in the detection image, combined with the detection image of positive control sample cells; secondary quantitative analysis of P-glycoprotein in the target biological sample cells within the 160-200 kDalton range in the detection image, combining protein glycosylation characteristics and detection images of target biological sample cells prepared by different lysis methods, compared with the detection image of positive control sample cells; and third-stage feature image learning of the detection images in the blurred drag region at the 150-300 kDalton position on the membrane, combined with the expression differences of P-glycoprotein in different tissue cells, to identify weak or no expression of P-glycoprotein in the target biological sample cells.

[0021] Based on the results of the three-level enhanced image learning, the risk of brain metastasis of solid tumors is comprehensively assessed.

[0022] Thirdly, embodiments of this application provide a computer device, the computer device including a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and, when executing the computer program, implement the method provided in any embodiment of this application.

[0023] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the method provided in any embodiment of this application.

[0024] This application employs a three-stage enhanced image learning approach, combining qualitative analysis of 141kDa baseline expression, quantitative analysis of glycosylation forms from 160-200kDa, and weak expression feature identification from 150-300kDa, covering different expression states of P-gp and addressing the issue of missed detection by traditional single-site detection methods. Samples are prepared using different lysis methods and compared to reduce detection bias caused by differences in cell lysis techniques. Combined with image enhancement processing and feature database matching, weak expression signals are effectively identified, avoiding false negatives. By integrating qualitative, quantitative, and feature learning results, a risk assessment model based on P-gp expression levels, glycosylation degree, and tissue specificity is constructed, providing quantifiable brain metastasis risk levels for clinical practice and guiding the development of individualized treatment plans.

[0025] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a schematic block diagram of a cell membrane protein pump efficiency monitoring system for assessing the risk of brain metastasis in solid tumors, provided in one embodiment of this application.

[0028] Figure 2 This is a schematic flowchart of the steps of a method for monitoring the efficiency of cell membrane protein pumps in assessing the risk of brain metastasis in solid tumors according to an embodiment of this application;

[0029] Figure 3 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0030] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0033] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0034] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0035] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0036] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0037] Brain metastasis of solid tumors is a significant prognostic factor for patients with advanced cancer. In particular, EGFR-mutant tumors such as non-small cell lung cancer (NSCLC) often result in insufficient drug concentrations penetrating the blood-brain barrier due to the drug efflux effect mediated by P-glycoprotein (P-gp), exacerbating drug resistance in brain metastases. As an ATP-dependent efflux pump, P-gp's glycosylation modification leads to changes in protein molecular weight, causing a shift in the protein band position in traditional Western blotting (WB) experiments. Furthermore, the expression of P-gp varies among different tissues and cells, easily leading to false-negative results due to weak or absent expression.

[0038] Current technologies for P-gp detection largely rely on qualitative analysis of single molecular weight sites, failing to consider variations in molecular weight range due to glycosylation modifications (e.g., 160-200 kDa glycosylation) and tissue-specific expression differences, making it difficult to accurately identify weakly expressed samples. Furthermore, traditional detection methods do not employ multi-fractionation sample preparation or segmented image learning strategies, failing to address the interference of glycosylation modifications and fragmentation methods on detection results, resulting in a lack of quantitative indicators for brain metastasis risk assessment. Although current technologies recognize the association between P-gp and brain metastasis, a systematic detection scheme targeting glycosylation modifications and weak expression characteristics has not yet been proposed, particularly lacking a third-order reinforcement learning method that incorporates multi-dimensional image analysis.

[0039] Therefore, there is an urgent need to design a technical solution to solve at least one of the above-mentioned technical problems.

[0040] To solve the above problem, please refer to Figure 1 This application provides a cell membrane protein pump efficiency monitoring system for assessing the risk of brain metastasis in solid tumors, comprising: a biological sample processing module for acquiring solid tumor biological sample cells, preparing test biological sample cells using different lysis methods, and acquiring positive control sample cells; an image acquisition module for performing protein detection on the test biological sample cells and positive control sample cells, acquiring detection images containing protein bands, wherein the protein detection includes detection of P-glycoprotein; and a step-by-step enhanced image learning module for performing three-order step-by-step enhanced image learning on the detection images, including: at a predicted 141 kDalton site in the detection image, combining the detection image of the positive control sample cells, detecting the P-glycoprotein of the test biological sample cells. Qualitative analysis of proteins is performed; within the 160-200 kDalton range of the detection image, the detection images of the test biological sample cells prepared by different lysis methods, combined with the protein glycosylation characteristics, are compared with the detection images of positive control sample cells to perform secondary quantitative analysis of P-glycoprotein in the test biological sample cells; in the blurred drag region appearing at the 150-300 kDalton position on the membrane in the detection image, combined with the expression differences of P-glycoprotein in different tissue cells, a third feature image learning is performed on the detection image to identify the weak expression or no expression of P-glycoprotein in the test biological sample cells; a risk assessment module is used to comprehensively assess the risk of brain metastasis of solid tumors based on the results of the three-stage enhanced image learning.

[0041] This system is designed for brain metastasis risk assessment in solid tumors (especially NSCLC). It focuses on monitoring the efficacy of cell membrane protein pumps (primarily P-gp). Through three-stage enhanced image learning, it addresses detection errors in traditional Western blotting (WB) experiments caused by P-gp glycosylation modifications and tissue expression differences. This enables precise qualitative and quantitative analysis of P-gp protein and identification of weak expression, ultimately quantifying the risk of brain metastasis. The core technology modules include: biological sample processing, image acquisition, enhanced image learning, and risk assessment. These modules are tightly coupled, forming a complete closed loop from sample preparation to risk assessment.

[0042] The biosample processing module is used to prepare test samples and positive control samples, preserving the glycosylation modification state and cell membrane integrity of P-gp proteins. Sample sources include primary / metastatic lesions from patients with solid tumors (such as NSCLC), circulating tumor cells (CTCs) in body fluids (cerebrospinal fluid, blood), or cancer cell samples obtained through cell line culture. Samples are processed using two or more lysis methods (such as mild nonionic detergent lysis buffer and strong denaturing lysis buffer): Mild lysis method: using lysis buffer containing NP-40 / Triton X-100 (supplemented with protease inhibitors) to preserve membrane protein integrity and glycosylation modification (avoiding glycan breakage), suitable for detecting native conformation P-gp. Strong lysis method: using lysis buffer containing SDS to destroy the higher-order structure of the protein and expose the core peptides, used to compare the effect of glycosylation on molecular weight. Positive control samples are selected from cell lines (such as KB-VIN, MDCK-MDR1 cells) or tissue samples that are known to highly express P-gp, and are simultaneously processed using the above lysis methods as a reference for image analysis.

[0043] The image acquisition module acquires detection images of P-gp protein bands through Western Blot (WB) experiments to ensure clear presentation of glycosylated bands.

[0044] The experimental procedures included: Protein separation: SDS-PAGE electrophoresis was performed on the lysed samples. The separating gel concentration was set to 8%-12% (covering the 141kDa range and the 160-200kDa range after glycosylation). Electrophoresis conditions: A constant voltage of 100V was applied until bromophenol blue entered the separating gel, then increased to 120V until the end to avoid degradation of glycosylated proteins due to high temperature. Transfer and blocking: A PVDF membrane (0.45μm pore size) was used for transfer at a constant current of 250mA for 90 minutes to ensure effective transfer of large glycosylated proteins. 5% skim milk powder was used as the blocking buffer (to avoid non-specific binding to glycosylation sites).

[0045] Antibody detection: Primary antibody: Specific antibodies recognizing the P-gp extracellular domain (such as C219, JCB129) are selected, covering both glycosylated and non-glycosylated epitopes. Secondary antibody: HRP-labeled goat anti-mouse / rabbit IgG, developed by chemiluminescence (ECL) with two exposure times: short (strong expression) and long (weak expression) to avoid signal saturation or loss. Image output: Grayscale images are acquired, including molecular weight markers (high-density labels in the 100-250 kDa range), the test sample lane, and the positive control lane, with a resolution ≥600 dpi to ensure band and drag details are discernible.

[0046] The phased enhancement image learning module addresses the qualitative, quantitative, and weak expression-related missed detection issues of P-gp detection through three-stage progressive image analysis.

[0047] First stage: Qualitative analysis of the 141kDa site (basic identification): Objective: To confirm the presence of non-glycosylated / core peptide form of P-gp (theoretical molecular weight 141kDa) in the sample. Based on the positive control lane, locate the 141kDa marker position and delineate a ±5kDa ROI (region of interest). Calculate the peak gray value of this region in the sample lane and compare it with the positive control threshold (preset ≥20% control peak value is positive) to determine the presence of basic expression (qualitative). Exclude non-specific bands: If a signal appears at the same position in the negative control (without primary antibody), activate the background subtraction algorithm.

[0048] Second-stage: Secondary quantification in the 160-200 kDa range (glycosylation correction): Objective: To correct for molecular weight shifts caused by glycosylation and quantify P-gp expression levels at different glycosylation degrees. Utilizing the migration characteristics of glycosylated proteins (molecular weight as a diffuse band in the 160-200 kDa range due to glycan length), multi-peak fitting (Gaussian mixture model) is performed within this range to identify glycosylation subtypes (high-mannose type, complex type). Comparison of lane differences between samples prepared using different lysis methods: The glycosylation band intensity of samples prepared using the mild lysis method should be higher than that prepared using the strong lysis method (because strong lysis destroys the glycan chain). If an inverse difference occurs, it indicates abnormal sample preparation. Quantitative calculation: Using the total area of ​​the glycosylated band in the positive control as a benchmark, the relative expression level of the sample to be tested = (sum of gray values ​​of the tested bands / sum of the positive control bands) × correction coefficient (lysis method difference coefficient).

[0049] Third Stage: Learning of Blurred Banding Features in the 150-300kDa Region (Weak Expression Recognition): Objective: To capture blurred signals caused by low abundance of P-gp or tissue-specific glycosylation, avoiding false negatives. Image enhancement algorithms (such as non-local mean denoising and adaptive histogram equalization) are used to highlight banding details in the 150-300kDa region. A tissue-specific expression database is constructed: weak expression patterns from samples from different sources such as lung cancer, liver cancer, and gastrointestinal tumors are included (e.g., faint banding around 180kDa is common in lung cancer samples). A convolutional neural network (CNN) is trained through transfer learning to identify common weak signal features across tissues. The "signal diffusion effect" of positive controls is introduced: if a positive control has banding in this region, and the test sample shows consecutive pixels with ≥10% control signal at the same location, it is judged as a weak expression positive.

[0050] Risk assessment module: Integrates the results of third-order image analysis to construct a quantitative model of brain metastasis risk.

[0051] Input parameters: Qualitative results (P-gp presence / absence); secondary quantitative values ​​(relative expression level of glycosylated proteins); weak expression identification results (positive / negative); clinical auxiliary data (primary tumor type, EGFR mutation status, blood-brain barrier integrity indicators). The evaluation model uses a weighted logistic regression model or random forest algorithm, with the following weight allocation: P-gp qualitative positivity (30%); glycosylated protein expression level ≥ 50% of positive controls (40%); weak expression identification positivity (20%); EGFR mutation status (10%).

[0052] Output results: Risk levels are divided into three levels: Low risk: no or weak P-gp expression and glycosylation level <30% (control); Medium risk: positive P-gp qualitative test and glycosylation level 30%-50% (control); High risk: strong P-gp expression (glycosylation level ≥50% (control)) or weak expression but accompanied by EGFR mutation.

[0053] By preserving protein modification states using different cleavage methods and combining this with molecular weight segmentation analysis, the limitations of traditional Western blotting (WB) based solely on molecular weight detection are overcome. For tissue-specific low-expression scenarios, image enhancement and transfer learning are utilized to reduce false negative rates. Integrating molecular detection and clinical data, quantifiable indicators of brain metastasis risk are provided, offering a basis for precise radiotherapy and drug dosage adjustment in EGFR-TKI-resistant patients.

[0054] Through the above implementation methods, the system has achieved standardization of the entire process from sample preparation to risk assessment, significantly improved the accuracy of P-gp protein detection, and filled the technical gap in the quantitative assessment of brain metastasis risk in solid tumors.

[0055] In some embodiments, obtaining solid tumor biological sample cells and preparing test biological sample cells using different lysis methods includes: isolating and obtaining biological sample cells from solid tumor patient samples, the samples including but not limited to cells derived from non-small cell lung cancer, liver cancer, and gastrointestinal tumors; selecting at least two different cell lysis methods to prepare test biological sample cells based on the characteristics of different tissue cells, the lysis methods including at least two of physical lysis, chemical lysis, or enzymatic lysis; and performing protein extraction and purification on the test biological sample cells prepared by different lysis methods to obtain a sample solution for detection.

[0056] This embodiment involves the acquisition of solid tumor biological sample cells and the processing of them using different lysis methods. The core is to preserve the protein modification state through various lysis methods based on the characteristics of cells from different tissue sources, so as to provide diverse samples for subsequent detection.

[0057] Sample separation involves obtaining biological sample cells from tissue biopsies, puncture samples, or body fluids (such as pleural effusion and ascites) of patients with solid tumors (e.g., non-small cell lung cancer, liver cancer, and gastrointestinal tumors) using enzymatic digestion (trypsin / collagenase) or density gradient centrifugation. For cultured tumor cell lines (e.g., A549 lung cancer cells and HepG2 liver cancer cells), cells in the logarithmic growth phase are collected after trypsin digestion.

[0058] Physical lysis method: Ultrasonic disruption (20kHz, 30s pulses x 3 times, ice bath cooling) or high-pressure homogenization is used. This method is suitable for cells with relatively tough cell membranes (such as liver cancer cells). During disruption, a cocktail of protease inhibitors (such as RocheComplete inhibitor) is added.

[0059] Chemical lysis methods include: Mild: using a lysis buffer containing 1% NP-40, 150mM NaCl, and 50mM Tris-HCl (pH 7.4), lysis for 30 min (4℃) to preserve glycosylation modifications of membrane proteins; Strong denaturing: using a lysis buffer containing 2% SDS and 100mM DTT, boiling for 5 min to destroy the higher-order structure of proteins, used to compare the effect of glycosylation on molecular weight. Enzymatic lysis: for samples rich in connective tissue, the extracellular matrix is ​​first digested with hyaluronidase, followed by chemical lysis.

[0060] After lysis, the samples were centrifuged at 12,000g for 15 min (4℃), and the supernatant was collected. The protein concentration was determined by the BCA method. Small molecule impurities were removed by ultrafiltration tube (10kDa molecular weight cutoff). The final protein concentration of the sample solution was adjusted to 2-5 μg / μL, and then aliquoted and frozen (-80℃).

[0061] In some embodiments, the step of detecting proteins in the test biological sample cells and positive control sample cells to obtain a detection image containing protein bands includes: separating the sample solutions of the test biological sample cells and the positive control sample cells by protein electrophoresis, allowing proteins to migrate in the gel according to their molecular weight; transferring the proteins in the gel to a solid membrane to form a protein membrane; then blocking the protein membrane, adding a specific antibody against P-glycoprotein for incubation, allowing the antibody to specifically bind to P-glycoprotein; and finally detecting the bound antibody by a colorimetric or fluorescent labeling method to obtain a detection image containing P-glycoprotein bands, wherein the detection image includes the distribution of protein bands corresponding to different lanes.

[0062] The detection images containing P-gp bands are obtained through steps such as protein electrophoresis, membrane transfer, and antibody incubation. The core is to ensure the effective separation of glycosylated proteins and the specificity of the signal.

[0063] Protein electrophoretic separation was performed by preparing a 10% SDS-PAGE gel (separating gel), with a sample loading volume of 20-30 μg / lane, and a molecular weight marker selected to cover 50-250 kDa (such as Thermo Scientific PageRulerPlus); electrophoresis conditions: constant voltage of 80V to the stacking gel, then increased to 120V (separating gel), and running until the bromophenol blue was 1 cm from the bottom of the gel to avoid excessive migration of glycosylated proteins.

[0064] The transfer and blocking were performed using a wet transfer method. After the PVDF membrane was activated with methanol, it was assembled with gel and filter paper into the transfer clamp and transferred at a constant current of 250mA for 90min (ice bath). After the transfer, the membrane was blocked with 5% skim milk powder (prepared with TBST) for 1h (room temperature) to block non-specific binding sites.

[0065] Primary antibody incubation was performed by adding P-gp-specific antibody (e.g., C219, diluted 1:1000) and shaking overnight at 4°C. Secondary antibody incubation was performed by incubating with HRP-labeled goat anti-mouse IgG (1:5000) at room temperature for 1 hour, followed by washing the membrane three times with TBST (10 min each time). Colorimetric development was performed by incubation with ECL chemiluminescence reagent for 2 min, followed by exposure using a ChemiDoc imaging system (short exposure of 10 s to detect strong signals, and long exposure of 1-5 min to capture weak signals). 8-bit grayscale images were saved, and lane information (sample number, lysis method type, and positive control) was labeled.

[0066] In some embodiments, the qualitative analysis of P-glycoprotein in the test biological sample cells at the predicted 141 kDalton site in the detection image, combined with the detection image of positive control sample cells, includes: locating a preset site region with a molecular weight of 141 kDalton in the detection image; extracting the band signal of the test biological sample cell lane within this site region and comparing it with the band signal of the positive control sample cell lane at the same site; if the test sample lane has a band at this site that matches the characteristics of the positive control signal, it is determined that the test sample contains the basic expression form of P-glycoprotein, thus completing the qualitative analysis.

[0067] By locating theoretical molecular weight sites and comparing them with positive control signals, the presence of non-glycosylated or core peptide forms of P-gp in the sample can be determined, thus achieving basic expression qualitative analysis.

[0068] Site localization and ROI delineation: Based on the molecular weight marker, locate the 141kDa±5kDa region (i.e., 136-146kDa) in the detection image, and set a rectangular ROI (region of interest) with a width covering a single lane.

[0069] Signal extraction and comparison were performed by extracting the average gray value within the ROI of the sample lane (background subtraction: average gray value of the area 2mm outside the ROI) and using the signal at the same position in the positive control lane as a benchmark. Signal feature matching included: band position (deviation from Marker < ±3kDa), shape (peak width < 10 pixels), and signal-to-noise ratio (S / N ≥ 2:1).

[0070] Qualitative determination: If the signal intensity of the test sample is ≥20% of the positive control signal and meets the feature matching condition, it is determined as "positive for P-gp basic expression"; if the signal intensity of the negative control (sample without primary antibody) at the same position is >10% of the positive control, the background subtraction algorithm (such as median filtering for noise reduction) is activated and the evaluation is re-evaluated.

[0071] In some embodiments, the secondary quantitative analysis of P-glycoprotein in the test biological sample cells, within the 160 to 200 kDalton range of the detection image, is performed by comparing the detection images of the test biological sample cells with those of the positive control sample cells, taking into account the protein glycosylation characteristics and different lysis methods. This includes: identifying band regions within the 160 to 200 kDalton molecular weight range in the detection image, since P-glycoprotein glycosylation may lead to an increase in molecular weight, the bands in this range reflect the protein form after glycosylation modification; extracting the band signal intensity and distribution characteristics of the test biological sample cells prepared by different lysis methods within this range, and comparing them with the band signals of the positive control sample cells in the same range using multiple lanes; determining the relative expression level of glycosylated P-glycoprotein in the test sample by calculating the ratio of the band signal intensity of the test sample to the signal intensity of the positive control, combined with the molecular weight shift law of glycosylation modification, thus completing the secondary quantitative analysis.

[0072] To address the increased molecular weight caused by glycosylation, the relative expression level of glycosylated P-gp was quantified by comparing samples from multiple cleavage methods, thus correcting the detection bias of the single cleavage method.

[0073] Band region identification involves defining a molecular weight range of 160-200 kDa in the detected image. This range exhibits diffuse bands (not single sharp bands) due to differences in sugar chain length, and edge detection algorithms (such as the Canny operator) are used to identify the band boundaries.

[0074] Multi-lane signal comparison is achieved by extracting the total gray value (integrated optical density, IOD) of samples from different lysis methods (e.g., mild / strong lysis method) within the specified range and normalizing it (IOD / protein loading amount). Positive control samples must provide both mild / strong lysis method lane signals to establish a glycosylation signal benchmark (the mild signal should be significantly higher than that of the strong lysis method, as strong lysis destroys the glycan chains).

[0075] Relative expression level calculation: Glycosylation expression level of the test sample = (mild IOD of the test sample / mild IOD of the positive control) × (strong lysis method IOD of the positive control / strong lysis method IOD of the test sample) × 100%; Introducing correction factor: If the strong lysis method signal of the test sample is abnormally elevated (indicating incomplete glycan destruction), multiply by the tissue-specific correction coefficient (e.g., 0.9 for lung cancer cells, 1.1 for liver cancer cells).

[0076] In some embodiments, the blurred banding region appearing at a position of 150 to 300 kilodaltons on the membrane in the detection image, combined with the expression differences of P-glycoprotein in different tissue cells, is used to perform a third feature image learning on the detection image to identify the weak or no expression of P-glycoprotein in the cells of the biological sample to be tested. This includes: performing image enhancement processing on the blurred banding region at a position of 150 to 300 kilodaltons in the detection image, including noise filtering, contrast adjustment, and edge detection; establishing a database of P-glycoprotein expression features of cells from different tissue sources, containing the banding morphology features of known weakly expressed or non-expressed samples; matching and comparing the processed features of the banding region of the sample to be tested with the features in the database; if a weak band or a banding signal of a specific morphology below a preset signal threshold is detected, and the signal appears in samples prepared by at least two lysis methods, it is determined to be weakly expressed P-glycoprotein; if no signal matching the features is detected, it is determined to be non-expressed, thus avoiding false negative results.

[0077] By using image enhancement and feature matching, we can identify blurred signals caused by low abundance or tissue-specific glycosylation, thus solving the problem of false negatives in traditional Western blotting.

[0078] Image enhancement processing involves nonlocal mean denoising of the 150-300kDa region (σ = 10, search window 21×21), adaptive histogram equalization (CLAHE) to improve contrast, and edge detection to highlight drag contours.

[0079] The feature database was established by collecting weakly expressed samples from different tissue sources (such as lung cancer brain metastases and colon cancer liver metastases) and labeling them with the following characteristics: location (e.g., 180kDa ± 20kDa), morphology (continuous ≥ 3 pixel blurred bands), and signal intensity (grayscale value < 5% of the positive control but > 2 times the background). A multi-dimensional feature vector was constructed, including band position offset, band width, and grayscale gradient change rate. Matching was performed by matching the processed sample's band characteristics with the database. If the following conditions were met: signal intensity > 2 times the background and < 15% of the positive control; bands appearing at the same location in at least two fragmentation methods; and the band morphology matching a certain type of tissue feature in the database with ≥ 70% accuracy, the sample was judged as "weakly expressed positive"; otherwise, it was judged as "no expression".

[0080] In some embodiments, the comprehensive assessment of the risk of brain metastasis of solid tumors based on the results of the three-stage enhanced image learning includes: integrating the presence of basic P-glycoprotein expression obtained from the first-stage qualitative analysis, the relative expression level of glycosylated proteins obtained from the second-stage quantitative analysis, and the identification results of weak expression or no expression obtained from the third-stage feature learning; weighting the analysis results of different stages according to a preset risk assessment model, wherein the expression level of glycosylated proteins and the identification results of weak expression are used as key risk factors; generating a risk level of brain metastasis of solid tumors based on the calculation results, wherein the risk level is correlated with the total expression level of P-glycoprotein, the degree of glycosylation modification, and tissue-specific expression differences, providing a quantitative basis for clinical assessment.

[0081] By integrating the results of the third-order analysis and combining them with a pre-set model for weighted calculation, a quantitative risk level is generated, providing a basis for clinical decision-making.

[0082] Data integration includes: First-order results: basal P-gp expression status (0 / 1, 1 for presence); Second-order results: relative glycosylation expression level (percentage, denoted as E%); Third-order results: weak expression status (0 / 1, 1 for weak expression); Clinical parameters: EGFR mutation status (0 / 1, 1 for mutation type), primary tumor type (lung cancer / liver cancer / gastrointestinal tumor, assigned different weights).

[0083] The weighted calculation model includes: Risk score = 0.4 × E% + 0.3 × (first-order result + third-order result) + 0.2 × EGFR mutation status + 0.1 × tissue risk factor (lung cancer = 1, liver cancer = 0.8, gastrointestinal tumors = 0.7); the tissue risk factor is based on literature data and reflects the tendency of different tumors to metastasize to the brain.

[0084] Risk levels are categorized as follows: Low risk: score < 40, E% < 30%, no weak expression; Intermediate risk: score 40 ≤ score < 70, or E% 30%-50%, or weak expression using a single cleavage method; High risk: score ≥ 70, or E% ≥ 50%, or weak expression using a double cleavage method + EGFR mutation. The output report includes the analysis results for each level, risk level, and clinical recommendations (e.g., enhanced brain MRI monitoring is recommended for high-risk patients).

[0085] In some embodiments, to address the limitations of manual feature extraction in traditional image analysis, an end-to-end multi-task deep learning model is designed to simultaneously perform qualitative and localization of P-gp protein, quantification of glycosylation, and identification of weak expression features. By sharing the underlying feature extraction layer and the task-specific output layer, detection efficiency and noise resistance are improved.

[0086] The model architecture design includes: Input layer: Normalized detection image (size 224×224×1, grayscale image), with lane position annotation information (XML coordinate file). Shared feature extraction layer: ResNet-34 is used as the backbone network, and the first 3 residual blocks extract general protein band features (edge, texture, grayscale distribution). Task branch layer: Qualitative branch: Fully connected layer after GlobalAverage Pooling, outputting the 141kDa site presence probability (binary classification, Sigmoid activation); Quantitative branch: Generates a heatmap in the 160-200kDa region, regressing the integrated optical density of glycosylated bands (Regression, MSE loss); Weak expression branch: Introduces an attention mechanism (SE module) to focus on the 150-300kDa drag region, outputting the weak expression probability (binary classification, combined with FocalLoss to handle class imbalance).

[0087] The training data construction included: annotation standards: 5000+ WB images were manually annotated, marking 141kDa site coordinates, glycosylation strip ROIs, and weak drag region contours; data augmentation: images were rotated (±15°), scaled (0.8-1.2 times), and Gaussian noise was added (σ=0.05) to alleviate the problem of insufficient sample size.

[0088] The reasoning process includes: after inputting the detection image, the model synchronously outputs: qualitative results: a probability of presence of 141kDa site ≥0.7 is judged as positive; quantitative values: peak value of glycosylation region heatmap × positive control calibration coefficient; weak expression probability: combined with the attention weight of the drag region, ≥0.6 is judged as weak expression positive.

[0089] In some embodiments, to address the differences in P-gp glycosylation patterns in cells from different tissue origins (e.g., lung cancer cells are predominantly of the high-mannose type, while liver cancer cells are predominantly of the complex type), a tissue-adaptive model based on transfer learning is constructed, and the generalization ability of weak signal detection across tissues is improved through a pre-training-fine-tuning strategy.

[0090] During the pre-training phase, ViT (VisionTransformer) is pre-trained on the publicly available WB image dataset (containing various protein bands) to learn the general visual features of protein electrophoresis images (such as band migration patterns and background noise patterns).

[0091] Tissue-specific fine-tuning includes: Grouping strategy: Samples are grouped according to tissue type (lung cancer / liver cancer / gastrointestinal tumors), and each group constructs a subset dataset (containing samples with strong expression, weak expression, and no expression of the corresponding tissue); Model adjustment: A tissue classification head (fully connected layer, outputting 3-class classification) and a weak signal detection head (binary classification) are connected after the ViT encoder, and jointly optimized through multi-task learning: L = αLweak + (1-α)Ltissue; where Lweak is the cross-entropy loss for weak signal detection, Ltissue is the tissue classification loss, and α = 0.8 prioritizes the optimization of weak signal detection. Detection implementation: When the image of the sample to be tested is input, the model first determines the source through the tissue classification head (probability ≥ 0.6 confirms the tissue type), and then calls the weak signal detection parameters of the corresponding tissue; For samples with unclear tissue source, an ensemble strategy is adopted: the prediction results of all tissue branches are fused, and the average probability ≥ 0.5 is used to determine the weak expression positive.

[0092] In some embodiments, a graph neural network model is constructed by integrating third-order image analysis results (qualitative / quantitative / weak expression), clinical data (EGFR mutation, tumor stage, blood-brain barrier permeability indicators) and protein interaction networks (association of P-gp with drug resistance proteins such as ABCB1 and MRP1) to capture the nonlinear relationships between complex factors.

[0093] The graph structure definition includes: Node types: Detection nodes: P-gp qualitative (0 / 1), glycosylation expression level (continuous value), weak expression (0 / 1); Clinical nodes: EGFR mutation (0 / 1), primary tumor type (category coding), age, PS score; Molecular nodes: ABCB1 gene polymorphism (rs1045642 locus genotype), MRP1 expression level (if detection data is available). Edge definition: Detection nodes and clinical nodes are connected by "association edges" (e.g., EGFR mutation → literature evidence of high P-gp expression), and molecular nodes and detection nodes are connected by "functional edges" (e.g., ABCB1 mutation → enhanced P-gp activity). The graph convolutional layer design uses GAT (Graph Attention Network) layers, which aggregate neighbor features for each node, and the weights are calculated by the attention mechanism (e.g., EGFR mutation nodes have higher attention weights for P-gp quantitative nodes); after two GAT layers, the node features are input into a fully connected layer, and the output is the brain metastasis risk probability (0-1).

[0094] Model training and application include: Data input: Constructing a graph structure for each patient's data, filling missing data (such as undetected MRP1) with the mean; Risk prediction: A probability ≥0.7 is considered high risk, 0.4-0.7 is medium risk, and <0.4 is low risk; Interpretability: Visualizing key influencing factors through attention weights (e.g., the edge weights for EGFR mutation and glycosylation expression are the highest in a patient, suggesting that the two synergistically increase risk).

[0095] In some embodiments, a CycleGAN variant is designed to address the common low-quality WB images (underexposure, blurred bands) in clinical practice, converting low signal-to-noise ratio images into high-quality, clear images, improving the success rate of weak signal recognition, and solving the problem of traditional methods' dependence on image quality.

[0096] GAN Architecture Design: Generator: U-Net structure, input low-quality image (containing noise, low contrast), output enhanced clear image (preserving real stripe details, suppressing artifacts); Discriminator: PatchGAN, determines whether a local image patch is a real high-quality image or a generated image, trained by combining pixel-level L1 loss and adversarial loss: L = LGAN + λL*L1 (λ = 100); Training data preparation includes: Paired dataset: low-quality images are generated by adding Gaussian noise (σ = 0.1-0.3) and reducing contrast (factor 0.5-0.8) to high-quality images, constructing 500 pairs of "low-quality-high-quality" image pairs; Unpaired data augmentation: CycleGAN's cycle consistency loss is introduced to ensure that the generated image can be restored back to the original low-quality image, avoiding mode collapse.

[0097] The detection process integration includes: first, the input detection image is quality assessed (based on the SSIM index < 0.6, it is judged as low quality); the low quality image is enhanced by the generator and then input into the subsequent third-order image analysis module (such as CNN model); experimental verification: in weak signal samples, the detection sensitivity is improved from 65% to 85% after GAN enhancement (compared with the traditional histogram equalization method).

[0098] In some embodiments, to address the needs of fragmented clinical data and privacy protection, a federated learning framework is constructed to train a risk assessment model using WB test data from multiple hospitals. This improves the model's generalization ability without sharing the original images and solves the problem of single-center data bias.

[0099] The federated learning architecture includes: Participants: Each hospital acts as a client, holding local testing data (image feature vectors + risk labels, excluding patient privacy information); Server: Coordinates model aggregation and uses the FedAvg algorithm to update global model parameters.

[0100] Feature extraction and privacy protection include: the client first preprocesses the detection image to extract intermediate features of the third-order analysis (such as 141kDa signal intensity, glycosylation region IOD, and weak drag feature vector), instead of the original image; the feature vector is uploaded after differential privacy processing (adding Laplacian noise, ε=0.5) to ensure data anonymization.

[0101] Model co-training includes: Initialization: The server distributes an initial risk assessment model (such as logistic regression or a lightweight neural network); Local training: The client trains the model using local feature data, and uploads gradient updates after 5-10 iterations; Global aggregation: The server weights and averages the parameters of each client (weights are proportional to the amount of data) to generate a new global model, with a cycle of once a week; Results: Cross-center validation shows that the AUC-ROC of the federated model is 12% higher than that of the single-center model, and it complies with GDPR data protection requirements.

[0102] It should be noted that the acquisition of any information mentioned in the system is in accordance with relevant regulations and with the user's consent, and will not infringe on the user's privacy or violate relevant laws and regulations.

[0103] Please see Figure 2 , Figure 2 This is a schematic flowchart of a method for monitoring the efficiency of cell membrane protein pumps in assessing the risk of brain metastasis in solid tumors, provided in one embodiment of this application. The device executing the method is a computer device deployed in the cell membrane protein pump efficiency monitoring system for assessing the risk of brain metastasis in solid tumors, provided in any embodiment of this application.

[0104] like Figure 2 As shown, the provided method includes steps S101 to S104. The computer device can be a handheld terminal, a laptop computer, a wearable device, or a robot, etc. This is used to implement steps S101 to S104 and their corresponding embodiments.

[0105] It should be noted that the acquisition of any information mentioned in the provided methods is in compliance with relevant regulations and is carried out with the user's consent, and will not infringe on the user's privacy or violate relevant laws and regulations.

[0106] Step S101. Obtain solid tumor biological sample cells, prepare test biological sample cells using different lysis methods, and obtain positive control sample cells.

[0107] Step S102. Perform protein detection on the biological sample cells to be tested and the positive control sample cells to obtain detection images containing protein bands, wherein the protein detection includes detection of P-glycoprotein.

[0108] Step S103. Perform three-stage enhanced image learning on the detection image, including: qualitative analysis of P-glycoprotein in the test biological sample cells at the predicted 141 kDalton site in the detection image, combined with the detection image of positive control sample cells; secondary quantitative analysis of P-glycoprotein in the test biological sample cells in the 160 to 200 kDalton range in the detection image, combined with the protein glycosylation characteristics and the detection images of test biological sample cells prepared by different lysis methods, compared with the detection images of positive control sample cells; and third-stage feature image learning on the detection image in the blurred drag region at the 150 to 300 kDalton position on the membrane in the detection image, combined with the expression differences of P-glycoprotein in different tissue cells, to identify the weak expression or no expression of P-glycoprotein in the test biological sample cells.

[0109] Step S104. Based on the results of the three-level enhanced image learning, comprehensively assess the risk of brain metastasis of solid tumors.

[0110] It should be noted that those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the cell membrane protein pump efficacy monitoring method and each step of the solid tumor brain metastasis risk assessment described above can be referred to the corresponding process in the embodiments of the cell membrane protein pump efficacy monitoring system for solid tumor brain metastasis risk assessment described above, and will not be repeated here.

[0111] Please see Figure 3 , Figure 3 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0112] The storage medium may store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform an embodiment of any method for monitoring the efficiency of cell membrane protein pumps in assessing the risk of brain metastasis in solid tumors.

[0113] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0114] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When executed by a processor, the computer program enables the processor to perform any method of monitoring the efficiency of cell membrane protein pumps for assessing the risk of brain metastasis in solid tumors.

[0115] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 3The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0116] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0117] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps:

[0118] Solid tumor biological sample cells were obtained, and test biological sample cells were prepared using different lysis methods. Positive control sample cells were also obtained.

[0119] Protein detection is performed on the biological sample cells to be tested and the positive control sample cells to obtain detection images containing protein bands. The protein detection includes the detection of P-glycoprotein.

[0120] The detection images undergo three-stage enhanced image learning, including: qualitative analysis of P-glycoprotein in the target biological sample cells at the predicted 141 kDalton site in the detection image, combined with the detection image of positive control sample cells; secondary quantitative analysis of P-glycoprotein in the target biological sample cells within the 160-200 kDalton range in the detection image, combining protein glycosylation characteristics and detection images of target biological sample cells prepared by different lysis methods, compared with the detection image of positive control sample cells; and third-stage feature image learning of the detection images in the blurred drag region at the 150-300 kDalton position on the membrane, combined with the expression differences of P-glycoprotein in different tissue cells, to identify weak or no expression of P-glycoprotein in the target biological sample cells.

[0121] Based on the results of the three-level enhanced image learning, the risk of brain metastasis of solid tumors is comprehensively assessed.

[0122] In some embodiments, obtaining solid tumor biological sample cells and preparing test biological sample cells using different lysis methods includes: isolating and obtaining biological sample cells from solid tumor patient samples, the samples including but not limited to cells derived from non-small cell lung cancer, liver cancer, and gastrointestinal tumors; selecting at least two different cell lysis methods to prepare test biological sample cells based on the characteristics of different tissue cells, the lysis methods including at least two of physical lysis, chemical lysis, or enzymatic lysis; and performing protein extraction and purification on the test biological sample cells prepared by different lysis methods to obtain a sample solution for detection.

[0123] In some embodiments, the step of detecting proteins in the test biological sample cells and positive control sample cells to obtain a detection image containing protein bands includes: separating the sample solutions of the test biological sample cells and the positive control sample cells by protein electrophoresis, allowing proteins to migrate in the gel according to their molecular weight; transferring the proteins in the gel to a solid membrane to form a protein membrane; then blocking the protein membrane, adding a specific antibody against P-glycoprotein for incubation, allowing the antibody to specifically bind to P-glycoprotein; and finally detecting the bound antibody by a colorimetric or fluorescent labeling method to obtain a detection image containing P-glycoprotein bands, wherein the detection image includes the distribution of protein bands corresponding to different lanes.

[0124] In some embodiments, the qualitative analysis of P-glycoprotein in the test biological sample cells at the predicted 141 kDalton site in the detection image, combined with the detection image of positive control sample cells, includes: locating a preset site region with a molecular weight of 141 kDalton in the detection image; extracting the band signal of the test biological sample cell lane within this site region and comparing it with the band signal of the positive control sample cell lane at the same site; if the test sample lane has a band at this site that matches the characteristics of the positive control signal, it is determined that the test sample contains the basic expression form of P-glycoprotein, thus completing the qualitative analysis.

[0125] In some embodiments, the secondary quantitative analysis of P-glycoprotein in the test biological sample cells, within the 160 to 200 kDalton range of the detection image, is performed by comparing the detection images of the test biological sample cells with those of the positive control sample cells, taking into account the protein glycosylation characteristics and different lysis methods. This includes: identifying band regions within the 160 to 200 kDalton molecular weight range in the detection image, since P-glycoprotein glycosylation may lead to an increase in molecular weight, the bands in this range reflect the protein form after glycosylation modification; extracting the band signal intensity and distribution characteristics of the test biological sample cells prepared by different lysis methods within this range, and comparing them with the band signals of the positive control sample cells in the same range using multiple lanes; determining the relative expression level of glycosylated P-glycoprotein in the test sample by calculating the ratio of the band signal intensity of the test sample to the signal intensity of the positive control, combined with the molecular weight shift law of glycosylation modification, thus completing the secondary quantitative analysis.

[0126] In some embodiments, the blurred banding region appearing at a position of 150 to 300 kilodaltons on the membrane in the detection image, combined with the expression differences of P-glycoprotein in different tissue cells, is used to perform a third feature image learning on the detection image to identify the weak or no expression of P-glycoprotein in the cells of the biological sample to be tested. This includes: performing image enhancement processing on the blurred banding region at a position of 150 to 300 kilodaltons in the detection image, including noise filtering, contrast adjustment, and edge detection; establishing a database of P-glycoprotein expression features of cells from different tissue sources, containing the banding morphology features of known weakly expressed or non-expressed samples; matching and comparing the processed features of the banding region of the sample to be tested with the features in the database; if a weak band or a banding signal of a specific morphology below a preset signal threshold is detected, and the signal appears in samples prepared by at least two lysis methods, it is determined to be weakly expressed P-glycoprotein; if no signal matching the features is detected, it is determined to be non-expressed, thus avoiding false negative results.

[0127] In some embodiments, the comprehensive assessment of the risk of brain metastasis of solid tumors based on the results of the three-stage enhanced image learning includes: integrating the presence of basic P-glycoprotein expression obtained from the first-stage qualitative analysis, the relative expression level of glycosylated proteins obtained from the second-stage quantitative analysis, and the identification results of weak expression or no expression obtained from the third-stage feature learning; weighting the analysis results of different stages according to a preset risk assessment model, wherein the expression level of glycosylated proteins and the identification results of weak expression are used as key risk factors; generating a risk level of brain metastasis of solid tumors based on the calculation results, wherein the risk level is correlated with the total expression level of P-glycoprotein, the degree of glycosylation modification, and tissue-specific expression differences, providing a quantitative basis for clinical assessment.

[0128] It should be noted that those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the processor described above can be referred to the corresponding process in the method embodiments of the above embodiments, and will not be repeated here.

[0129] The embodiments of this application also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, and the processor executing the program instructions to implement the steps of the cell membrane protein pump efficiency monitoring method for solid tumor brain metastasis risk assessment provided in the above embodiments of this application.

[0130] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0131] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A cell membrane protein pump efficiency monitoring system for assessing the risk of brain metastasis in solid tumors, characterized in that, include: The biological sample processing module is used to obtain solid tumor biological sample cells, prepare test biological sample cells using different lysis methods, and obtain positive control sample cells. The image acquisition module is used to perform protein detection on the biological sample cells to be tested and the positive control sample cells, and to acquire detection images containing protein bands, wherein the protein detection includes the detection of P-glycoprotein; The phased enhancement image learning module is used to perform three-stage phased enhancement image learning on the detection image, including: qualitative analysis of P-glycoprotein in the target biological sample cells at the predicted 141 kDalton site in the detection image, combined with the detection image of positive control sample cells; secondary quantitative analysis of P-glycoprotein in the target biological sample cells in the 160 to 200 kDalton range in the detection image, combined with the protein glycosylation characteristics and the detection images of the target biological sample cells prepared by different lysis methods, compared with the detection images of positive control sample cells; and third feature image learning on the detection image in the blurred drag region at the 150 to 300 kDalton position on the membrane, combined with the expression differences of P-glycoprotein in different tissue cells, to identify the weak expression or no expression of P-glycoprotein in the target biological sample cells. The risk assessment module is used to comprehensively assess the risk of brain metastasis of solid tumors based on the results of the three-level enhanced image learning.

2. The system according to claim 1, characterized in that, The acquisition of solid tumor biological sample cells involves preparing the test biological sample cells using different lysis methods, including: Biological sample cells are isolated and obtained from samples of patients with solid tumors, including but not limited to cells derived from non-small cell lung cancer, liver cancer, and gastrointestinal tumors; Based on the characteristics of different tissue cells, at least two different cell lysis methods are selected to prepare biological sample cells for testing, wherein the lysis methods include at least two of the following: physical lysis, chemical lysis, or enzymatic lysis. Proteins were extracted and purified from the biological sample cells prepared by different lysis methods to obtain sample solutions for detection.

3. The system according to claim 1, characterized in that, The process of performing protein detection on the biological sample cells to be tested and the positive control sample cells to obtain detection images containing protein bands includes: The sample solutions of the biological sample cells to be tested and the sample solutions of the positive control sample cells were separated by protein electrophoresis, so that the proteins migrate in the gel according to their molecular weight. The protein in the gel is transferred to a solid membrane to form a protein membrane. The protein membrane is then blocked and incubated with a specific antibody against P-glycoprotein to allow the antibody to bind specifically to P-glycoprotein. Finally, the bound antibody is detected by colorimetric or fluorescent labeling methods to obtain a detection image containing P-glycoprotein bands. The detection image includes the distribution of protein bands corresponding to different lanes.

4. The system according to claim 1, characterized in that, The qualitative analysis of P-glycoprotein in the tested biological sample cells is performed at the predicted 141 kDalton site in the detection image, combined with the detection image of the positive control sample cells, including: Locate the preset site region with a molecular weight of 141 kilodaltons in the detection image; The band signal of the cell lane of the biological sample to be tested in the region of the site is extracted and compared with the band signal of the cell lane of the positive control sample at the same site; If a band matching the signal characteristics of the positive control is found at this site in the lane of the sample to be tested, it is determined that the basic expression form of P-glycoprotein exists in the sample to be tested, and the qualitative analysis is completed.

5. The system according to claim 1, characterized in that, The detection images of the test biological sample cells, within the range of 160 to 200 kilodaltons in the detection image, combined with the protein glycosylation characteristics and different lysis methods, are compared with the detection images of positive control sample cells to perform a secondary quantitative analysis of the P-glycoprotein in the test biological sample cells, including: Identify and detect band regions in the molecular weight range of 160 to 200 kilodaltons in the detection image. Since P-glycoprotein glycosylation may lead to an increase in molecular weight, the bands in this range reflect the protein form after glycosylation modification. The intensity and distribution characteristics of band signals in biological sample cells prepared by different lysis methods within this range were extracted and compared with the band signals in the same range of positive control sample cells in multiple lanes. By calculating the ratio of the signal intensity of the test sample to the signal intensity of the positive control, and combining the molecular weight shift pattern of glycosylation modification, the relative expression level of glycosylated P-glycoprotein in the test sample was determined, and a secondary quantitative analysis was completed.

6. The system according to claim 1, characterized in that, The blurred, trailing region appearing at a position of 150 to 300 kilodaltons on the membrane in the detection image, combined with the expression differences of P-glycoprotein in different tissue cells, is used to perform a third feature image learning on the detection image to identify weak or no expression of P-glycoprotein in the cells of the biological sample under test, including: Image enhancement processing is performed on the blurred drag regions in the 150 to 300 kDalton range of the detected image, including noise filtering, contrast adjustment and edge detection; Establish a database of P-glycoprotein expression characteristics in cells from different tissue sources, including the drag morphological characteristics of known weakly expressed or non-expressed samples; The drag region features of the processed test sample are matched and compared with the features in the database. If a weak band or a drag signal of a specific shape below the preset signal threshold is detected, and the signal appears in samples prepared by at least two lysis methods, it is determined to be weak expression of P-glycoprotein. If no signal matching the characteristics is detected, it is determined to be no expression, thus avoiding false negative results.

7. The system according to claim 1, characterized in that, The comprehensive assessment of the risk of brain metastasis of solid tumors based on the results of the three-level enhanced image learning includes: The study integrates the basic expression of P-glycoprotein obtained from the first-order qualitative analysis, the relative expression level of glycosylated proteins obtained from the second-order quantitative analysis, and the weak expression or no expression identification results obtained from the third-order feature learning. Based on the pre-set risk assessment model, the analysis results of different levels are weighted and calculated, with the expression level of glycosylated proteins and the weak expression identification results serving as key risk factors. Based on the calculation results, a risk level for brain metastasis of solid tumors is generated. The risk level is associated with the total expression level of P-glycoprotein, the degree of glycosylation modification, and tissue-specific expression differences, providing a quantitative basis for clinical assessment.

8. A method for monitoring cell membrane protein pump efficiency in assessing the risk of brain metastasis in solid tumors, characterized in that, The cell membrane protein pump efficiency monitoring system for assessing the risk of brain metastasis in solid tumors according to any one of claims 1-7, the method comprising: Solid tumor biological sample cells were obtained, and test biological sample cells were prepared using different lysis methods. Positive control sample cells were also obtained. Protein detection is performed on the biological sample cells to be tested and the positive control sample cells to obtain detection images containing protein bands. The protein detection includes the detection of P-glycoprotein. The detection images undergo three-stage enhanced image learning, including: qualitative analysis of P-glycoprotein in the target biological sample cells at the predicted 141 kDalton site in the detection image, combined with the detection image of positive control sample cells; secondary quantitative analysis of P-glycoprotein in the target biological sample cells within the 160-200 kDalton range in the detection image, combining protein glycosylation characteristics and detection images of target biological sample cells prepared by different lysis methods, compared with the detection image of positive control sample cells; and third-stage feature image learning of the detection images in the blurred drag region at the 150-300 kDalton position on the membrane, combined with the expression differences of P-glycoprotein in different tissue cells, to identify weak or no expression of P-glycoprotein in the target biological sample cells. Based on the results of the three-level enhanced image learning, the risk of brain metastasis of solid tumors is comprehensively assessed.

9. A computer device, characterized in that, The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and, in executing the computer program, implement the method as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, causes the processor to implement the method as described in claim 8.

Images