A method for analyzing the surface glycosylation profile of plasma VWF using lectin chip

Abstract

The present invention discloses a method for analyzing the surface glycosylation spectrum of plasma VWF using a lectin chip, comprising the following steps: using an effective sample recognition model pre-established based on a neural network model to screen out effective samples of healthy blood samples and effective samples of low VWF blood samples for analyzing specific VWF glycosylation spectrum, and performing a lectin binding comparison analysis on effective samples of healthy blood samples and effective samples of low VWF blood samples to obtain specific lectins that bind to low VWF blood samples, and using the lectin binding forces of specific lectins with healthy blood samples and low VWF blood samples to draw a VWF glycosylation spectrum that characterizes the surface glycosylation characteristics of blood sample VWF. The present invention adds machine vision, and uses a neural network model for artificial intelligence automatic identification to screen out effective samples of specific VWF glycosylation spectrum from all experimental samples, so as to avoid interference of invalid samples with subsequent analysis of VWF glycosylation spectrum.

Description

Technical Field

[0001] The present invention relates to the technical field of glycosylation analysis, and in particular to a method for analyzing the surface glycosylation of plasma VWF using a lectin chip. Background Art

[0002] Von Willebrand factor (VWF) is a highly sialylated glycoprotein synthesized by vascular endothelial cells and megakaryocytes / platelets. VWF plays two important roles in normal hemostasis: one is to bind to the platelet membrane glycoprotein (GP) Ib-IX complex to mediate platelet adhesion to the exposed subendothelial tissue at the site of vascular injury; the other is to serve as a carrier molecule for coagulation factor VIII (factor VIII, FVIII). A portion of the VWF synthesized by vascular endothelial cells is directly secreted into the blood circulation, while most of it is stored in Weibel-Palade bodies. VWF in the blood circulation mainly comes from endothelial cells. The VWF protein monomer contains 13 N-type and 10 O-type glycans (N-and O-glycans) sites, of which more than 90% of the terminal sugars of N- and O-type sugars such as galactose (Gal) or N-acetylgalactosamine (GalNAc) are further modified by sialic acid in the Golgi apparatus. Sialylation of VWF glycans plays a key role in regulating its clearance.

[0003] Glycan determinants on von Willebrand factor (VWF) play a role in regulating its susceptibility to proteolysis and clearance. Abnormal glycosylation has been shown to cause von Willebrand disease (VWD) in many different mouse models. However, accurate assessment of VWF glycan properties is not possible with existing techniques because the importance of glycan composition carbohydrates in the pathogenesis of human VWD remains largely unexplored. Summary of the invention

[0004] The purpose of the present invention is to provide a method for analyzing the surface glycosylation of plasma VWF using a VWF lectin chip, so as to solve the technical problem that the VWF glycan characteristics cannot be accurately evaluated in the prior art.

[0005] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0006] A method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip comprises the following steps:

[0007] Step S1, selecting a group of healthy persons and a group of low VWF patient plasma samples as a group of healthy blood samples and a group of low VWF blood samples respectively, using anti-VWF polyclonal antibody coated ELISA to detect VWF:Ag concentration, and diluting each healthy blood sample and each low VWF blood sample in 0.1% PBS-Tween so that the VWF:Ag concentration in each healthy blood sample and each low VWF blood sample reaches 1 μg / mL;

[0008] Step S2, incubating each healthy blood sample and each low VWF blood sample with the lectin chip for 1 hour, obtaining a lectin dot matrix fluorescence image of the lectin chip, using the pre-established chip validity judgment model to determine the chip validity of each lectin chip using the lectin dot matrix fluorescence image, and then screening out effective lectin chips from all lectin chips based on the chip validity, and using a chip scanner to read out the lectin binding force of each lectin in the effective lectin chip with the healthy blood sample and the low VWF blood sample after incubating the effective lectin chip for 1 hour;

[0009] Step S3, using the effective sample recognition model pre-established based on the neural network model to automatically identify the binding force images of each lectin in the effective lectin chip with the lectin binding force of the healthy blood sample and the low VWF blood sample, so as to screen out the effective samples of healthy blood samples and the effective samples of low VWF blood samples for analyzing the specific VWF glycosylation spectrum from a group of healthy blood samples and a group of low VWF blood samples, and perform lectin binding force comparison analysis on the effective samples of healthy blood samples and the effective samples of low VWF blood samples to obtain the specific lectin binding to the low VWF blood sample, and draw the VWF glycosylation spectrum characterizing the surface glycosylation characteristics of VWF in the blood sample based on the binding force of the specific lectin with the lectin of the healthy blood sample and the low VWF blood sample.

[0010] As a preferred embodiment of the present invention, the chip validity of each lectin chip is determined by using the lectin dot matrix fluorescence image using a pre-established chip validity determination model, including:

[0011] The lectin dot matrix fluorescence image of each healthy blood sample is sequentially input into the chip validity determination model, and the chip validity determination model outputs the chip validity of the lectin dot matrix fluorescence image of each healthy blood sample;

[0012] The lectin dot matrix fluorescence image of each low VWF blood sample is sequentially input into the chip validity determination model, and the chip validity determination model outputs the chip validity of the lectin dot matrix fluorescence image of each low VWF blood sample;

[0013] A set of defective historical lectin dot matrix fluorescence incomplete images is obtained, and all the historical lectin dot matrix fluorescence incomplete images are manually marked with chip validity category labels, wherein the chip validity category labels include incomplete images and non-incomplete images;

[0014] Using the historical lectin dot matrix fluorescence defect image as the first input item of the softmax neural network, using the chip validity category label of the historical lectin dot matrix fluorescence defect image as the first output item of the softmax neural network, and performing convolution training on the softmax neural network based on the first input item and the first output item to obtain the chip validity determination model;

[0015] The model expression of the chip validity determination model is:

[0016] Label1=softmax(Figure_Vector1);

[0017] Where Label1 is the expression identifier of the chip validity category label, softmax is the expression identifier of the softmax neural network, and Figure_Vector1 is the expression identifier of the lectin dot matrix fluorescence defect image.

[0018] As a preferred embodiment of the present invention, the method of screening out effective lectin chips from all lectin chips based on chip effectiveness comprises:

[0019] The lectin chips corresponding to the lectin dot matrix fluorescence defective images with chip validity category labels of non-defective in all healthy blood samples are screened out and marked as valid lectin chips for healthy blood samples;

[0020] The lectin chips corresponding to the lectin dot matrix fluorescence defect images with chip validity category label as non-defective in all low VWF blood samples were screened out and marked as valid lectin chips for low VWF blood samples.

[0021] As a preferred embodiment of the present invention, the method of automatically identifying the binding force images of the binding forces of each lectin in the effective lectin chip with the lectin of healthy blood samples and low VWF blood samples by using the effective sample recognition model pre-established based on the neural network model comprises:

[0022] The binding force image of each healthy blood sample is sequentially input into a healthy blood sample valid sample identification model, and the healthy blood sample valid sample identification model outputs the sample validity of each healthy blood sample;

[0023] sequentially inputting the binding force image of each low VWF blood sample into a low VWF blood sample valid sample identification model, and the low VWF blood sample valid sample identification model outputs the sample validity of each low VWF blood sample;

[0024] The construction of the effective sample identification model of healthy blood samples includes:

[0025] Acquire a group of binding force images of historical healthy blood samples, and manually mark the binding force images of all historical healthy blood samples with sample validity category labels, wherein the sample validity category labels include invalid blood samples and valid blood samples;

[0026] Using the binding force image of the historical healthy blood sample as the second input item of the softmax neural network, using the sample validity category label of the binding force image of the historical healthy blood sample as the second output item of the softmax neural network, and performing convolution training on the softmax neural network based on the second input item and the second output item to obtain the healthy blood sample valid sample recognition model;

[0027] The model expression of the healthy blood sample effective sample identification model is:

[0028] Labe2=softmax(Figure_Vector2);

[0029] Wherein, Labe2 is the expression identifier of the validity category label of the healthy blood sample, softmax is the expression identifier of the softmax neural network, and Figure_Vector2 is the expression identifier of the binding force image of the healthy blood sample;

[0030] The construction of the effective sample identification model for low VWF blood samples includes:

[0031] Acquire a group of binding force images of historically low VWF blood samples, and manually mark the binding force images of all historically low VWF blood samples with sample validity category labels, wherein the sample validity category labels include invalid blood samples and valid blood samples;

[0032] Using the binding force image of the historical low VWF blood sample as the third input item of the softmax neural network, using the sample validity category label of the binding force image of the historical low VWF blood sample as the third output item of the softmax neural network, and performing convolution training on the softmax neural network based on the third input item and the third output item to obtain the low VWF blood sample valid sample recognition model;

[0033] The model expression of the low VWF blood sample effective sample identification model is:

[0034] Label3=softmax(Figure_Vector3);

[0035] Where Label3 is the expression identifier of the low VWF sample validity category label, softmax is the expression identifier of the softmax neural network, and Figure_Vector3 is the expression identifier of the low VWF binding image.

[0036] As a preferred embodiment of the present invention, the screening of effective samples of healthy blood samples and effective low VWF blood samples includes:

[0037] Healthy blood samples with a sample validity category label of valid blood samples are screened out from all healthy blood samples and marked as valid samples of healthy blood samples;

[0038] The low VWF blood samples with the sample validity category label as valid blood samples among all the low VWF blood samples are screened out and marked as valid samples of low VWF blood samples.

[0039] As a preferred embodiment of the present invention, obtaining the specific lectin comprises:

[0040] n valid samples are selected from both the valid samples of healthy blood samples and the valid samples of low VWF blood samples;

[0041] The lectin binding force of each lectin in the valid samples of n healthy blood samples is summarized to obtain the lectin binding force vector of each lectin in the healthy blood sample. The function expression of the lectin binding force vector in the healthy blood sample is: X i ={ x ij | i ∈[1,m], j ∈[1,n]}, where X i For healthy blood samples i The lectin binding force vector of each lectin, x ij For healthy blood samples j The valid samples and i The lectin binding capacity of each lectin, m is the total number of lectins in the effective lectin chip;

[0042] The lectin binding force of each lectin in the n valid samples of low VWF blood samples is summarized in turn to obtain the lectin binding force vector of each lectin in the low VWF blood sample. The function expression of the lectin binding force vector in the low VWF blood sample is: Y i ={ y ij | i ∈[1,m], j ∈[1,n]}, where Y i The first i The lectin binding force vector of each lectin, yij The first j The valid samples and i The lectin binding capacity of each lectin, m is the total number of lectins in the effective lectin chip;

[0043] Calculate the lectin binding force vector X in healthy blood samples i Binding force vector Y to lectin in low VWF blood samples i and compare the similarity with the similarity threshold, wherein when the similarity is greater than the similarity threshold, the first i The first lectin is a specific lectin;

[0044] When the similarity is less than or equal to the similarity threshold, the first i The lectin is a non-specific lectin;

[0045] The similarity is measured using Euclidean distance, and the similarity calculation formula is: Where, P i For X i and Y i similarity, T is the transposition operator.

[0046] As a preferred embodiment of the present invention, the method of drawing a VWF glycosylation profile characterizing the surface glycosylation characteristics of VWF in the blood sample by combining the binding force of the specific lectin with the lectin of the healthy blood sample and the low VWF blood sample comprises:

[0047] The lectin binding force corresponding to the specific lectin extracted from the effective sample of the healthy blood sample is plotted to form the VWF glycosylation profile of the healthy blood sample representing the surface glycosylation characteristics of VWF in the healthy blood sample;

[0048] The lectin binding force corresponding to the specific lectin extracted from the effective sample of the low VWF blood sample was plotted to form the VWF glycosylation profile of the low VWF blood sample to characterize the VWF surface glycosylation characteristics of the low VWF blood sample.

[0049] As a preferred embodiment of the present invention, the lectin chip includes concanavalin A, snowdrop agglutinin, bean white agglutinin, bean red agglutinin, interstitial mandshurica agglutinin, crystal agglutinin, purpura leukohemagglutinin, amorpha fruticosa agglutinin, black ginseng agglutinin, leukohemagglutinin, euclid agglutinin-I, lotus root agglutinin, ricinus communis agglutinin-I, crystal erythrina agglutinin, peanut agglutinin, japanese arborvitae, tremella fuciformis agglutinin IB4, wisteria vine agglutinin, soybean agglutinin, spiral pomace agglutinin, amomum villosum agglutinin, biflorus duric lectin, tremella fuciformis agglutinin II, wheat germ agglutinin, tomato agglutinin and potato agglutinin. As a preferred embodiment of the present invention, the specific lectin includes black ginseng agglutinin and ricinus communis agglutinin-I.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The present invention incorporates machine vision and utilizes a neural network model for artificial intelligence automated recognition, thereby screening out effective samples of specific VWF glycosylation profiles from all experimental samples, thereby preventing invalid samples from interfering with subsequent analysis of the VWF glycosylation profile, and performing a blood sample control analysis to obtain specific lectins for accurately evaluating VWF glycan characteristics, and subsequently only needs to perform agglutination analysis of black ginseng lectin and ricinus communis lectin-I to obtain VWF glycan characteristics characterizing the blood sample, thereby improving the analysis efficiency and accuracy of plasma VWF surface glycosylation analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0053] Figure 1 This is a flow chart of the VWF surface glycosylation analysis method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] like Figure 1As shown, the present invention provides a method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip, comprising the following steps:

[0056] A method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip comprises the following steps:

[0057] Step S1, selecting a group of healthy persons and a group of low VWF patient plasma samples as a group of healthy blood samples and a group of low VWF blood samples respectively, using anti-VWF polyclonal antibody coated ELISA to detect VWF:Ag concentration, and diluting each healthy blood sample and each low VWF blood sample in 0.1% PBS-Tween so that the VWF:Ag concentration in each healthy blood sample and each low VWF blood sample reaches 1 μg / mL, and both the healthy blood sample and the low VWF blood sample contain fluorescent markers;

[0058] Step S2, incubating each healthy blood sample and each low VWF blood sample with the lectin chip for 1 hour, obtaining a lectin dot matrix fluorescence image of the lectin chip, using the pre-established chip validity judgment model to determine the chip validity of each lectin chip using the lectin dot matrix fluorescence image, and then screening out effective lectin chips from all lectin chips based on the chip validity, and using a chip scanner to read out the lectin binding force of each lectin in the effective lectin chip with the healthy blood sample and the low VWF blood sample after incubating the effective lectin chip for 1 hour;

[0059] The lectin dot matrix fluorescence images are used to determine the chip validity of each lectin chip using a pre-established chip validity judgment model, including:

[0060] The lectin dot matrix fluorescence image of each healthy blood sample is sequentially input into the chip validity determination model, and the chip validity determination model outputs the chip validity of the lectin dot matrix fluorescence image of each healthy blood sample;

[0061] The lectin dot matrix fluorescence image of each low VWF blood sample is sequentially input into the chip validity determination model, and the chip validity determination model outputs the chip validity of the lectin dot matrix fluorescence image of each low VWF blood sample;

[0062] A set of defective historical lectin dot matrix fluorescence incomplete images are obtained, and all the historical lectin dot matrix fluorescence incomplete images are manually labeled with chip validity category labels, where the chip validity category labels include incomplete images and non-incomplete images;

[0063] The historical lectin dot matrix fluorescence defect image is used as the first input item of the softmax neural network, the chip validity category label of the historical lectin dot matrix fluorescence defect image is used as the first output item of the softmax neural network, and the softmax neural network is convolutionally trained based on the first input item and the first output item to obtain a chip validity determination model;

[0064] The model expression of the chip validity determination model is:

[0065] Label1=softmax(Figure_Vector1);

[0066] Where Label1 is the expression identifier of the chip validity category label, softmax is the expression identifier of the softmax neural network, and Figure_Vector1 is the expression identifier of the lectin dot matrix fluorescence defect image.

[0067] Based on the chip effectiveness, effective lectin chips were screened out from all lectin chips, including:

[0068] The lectin chips corresponding to the lectin dot matrix fluorescence defective images with chip validity category labels of non-defective in all healthy blood samples are screened out and marked as valid lectin chips for healthy blood samples;

[0069] The lectin chips corresponding to the lectin dot matrix fluorescence defect images with chip validity category label as non-defective in all low VWF blood samples were screened out and marked as valid lectin chips for low VWF blood samples.

[0070] By building a model to screen the chip validity, incomplete lectin chips can be automatically eliminated to avoid entering the subsequent process and affecting the accuracy of subsequent analysis results. In addition, compared with manual identification, automatic identification reduces the burden on personnel and avoids the subjectivity of manual identification.

[0071] Step S3, using the effective sample recognition model pre-established based on the neural network model to automatically identify the binding force images of each lectin in the effective lectin chip with the lectin binding force of the healthy blood sample and the low VWF blood sample, so as to screen out the effective samples of healthy blood samples and the effective samples of low VWF blood samples for analyzing the specific VWF glycosylation spectrum from a group of healthy blood samples and a group of low VWF blood samples, and perform lectin binding force comparison analysis on the effective samples of healthy blood samples and the effective samples of low VWF blood samples to obtain the specific lectin binding to the low VWF blood sample, and draw the VWF glycosylation spectrum characterizing the surface glycosylation characteristics of VWF in the blood sample based on the binding force of the specific lectin with the lectin of the healthy blood sample and the low VWF blood sample.

[0072] The effective sample recognition model pre-established based on the neural network model is used to automatically identify the binding force images of each lectin in the effective lectin chip with the lectin binding force of healthy blood samples and low VWF blood samples, including:

[0073] The binding force image of each healthy blood sample is sequentially input into a healthy blood sample valid sample identification model, and the healthy blood sample valid sample identification model outputs the sample validity of each healthy blood sample;

[0074] sequentially inputting the binding force image of each low VWF blood sample into a low VWF blood sample valid sample identification model, and the low VWF blood sample valid sample identification model outputs the sample validity of each low VWF blood sample;

[0075] The construction of the effective sample identification model for healthy blood samples includes:

[0076] A set of binding force images of historical healthy blood samples are obtained, and the binding force images of all historical healthy blood samples are manually labeled with sample validity category labels, where the sample validity category labels include invalid blood samples and valid blood samples;

[0077] The binding force image of the historical healthy blood sample is used as the second input item of the softmax neural network, the sample validity category label of the binding force image of the historical healthy blood sample is used as the second output item of the softmax neural network, and the softmax neural network is convolutionally trained based on the second input item and the second output item to obtain a healthy blood sample valid sample recognition model;

[0078] The model expression of the healthy blood sample effective sample recognition model is: Labe2=softmax(Figure_Vector2);

[0079] Wherein, Labe2 is the expression identifier of the validity category label of the healthy blood sample, softmax is the expression identifier of the softmax neural network, and Figure_Vector2 is the expression identifier of the binding force image of the healthy blood sample;

[0080] The construction of an effective sample identification model for low VWF blood samples includes:

[0081] A set of binding force images of historically low VWF blood samples are obtained, and all binding force images of historically low VWF blood samples are manually labeled with sample validity category labels, where the sample validity category labels include invalid blood samples and valid blood samples;

[0082] Using the binding force image of the historical low VWF blood sample as the third input item of the softmax neural network, using the sample validity category label of the binding force image of the historical low VWF blood sample as the third output item of the softmax neural network, and performing convolution training on the softmax neural network based on the third input item and the third output item to obtain a low VWF blood sample valid sample recognition model;

[0083] The model expression of the effective sample recognition model for low VWF blood samples is:

[0084] Label3=softmax(Figure_Vector3); where Label3 is the expression identifier of the low VWF sample validity category label, softmax is the expression identifier of the softmax neural network, and Figure_Vector3 is the expression identifier of the low VWF binding image. The model is constructed for sample validity screening, which can automatically eliminate unqualified blood samples to avoid entering the subsequent process and affecting the accuracy of subsequent analysis results. In addition, compared with manual identification, automatic identification reduces the burden on personnel and avoids the subjectivity of manual identification.

[0085] Screening of valid samples of healthy blood samples and valid low VWF blood samples, including:

[0086] Healthy blood samples with a sample validity category label of valid blood samples are screened out from all healthy blood samples and marked as valid samples of healthy blood samples;

[0087] The low VWF blood samples with the sample validity category label as valid blood samples among all the low VWF blood samples are screened out and marked as valid samples of low VWF blood samples.

[0088] The acquisition of specific lectins includes:

[0089] n valid samples are selected from both the valid samples of healthy blood samples and the valid samples of low VWF blood samples;

[0090] The lectin binding force of each lectin in the valid samples of n healthy blood samples is summarized to obtain the lectin binding force vector of each lectin in the healthy blood sample. The function expression of the lectin binding force vector in the healthy blood sample is: X i ={ x ij | i ∈[1,m], j ∈[1,n]}, where X i For healthy blood samples i The lectin binding force vector of each lectin, x ij For healthy blood samples j The valid samples and iThe lectin binding force of n lectins, m is the total number of lectins in the effective lectin chip; the lectin binding force of each lectin in the effective samples of n low VWF blood samples is summarized in turn to obtain the lectin binding force vector of each lectin in the low VWF blood sample. The function expression of the lectin binding force vector in the low VWF blood sample is: i ={ y ij | i ∈[1,m], j ∈[1,n]}, where Y i The first i The lectin binding force vector of each lectin, y ij The first j The valid samples and i The lectin binding capacity of each lectin, m is the total number of lectins in the effective lectin chip;

[0091] Calculate the lectin binding force vector X in healthy blood samples i Binding force vector Y to lectin in low VWF blood samples i and compare the similarity with the similarity threshold, where

[0092] When the similarity is greater than the similarity threshold, the first i The first lectin is a specific lectin;

[0093] When the similarity is less than or equal to the similarity threshold, the first i The lectin is a non-specific lectin;

[0094] The similarity is measured using Euclidean distance, and the similarity calculation formula is: Where, P i For X i and Y i similarity, T is the transposition operator.

[0095] For i For each lectin, the lower the similarity between a group of low VWF blood samples and a group of healthy blood samples, the more i The low VWF blood sample on the first lectin has a specific performance, that is, the subsequent i The VWF glycan profiles of blood samples were characterized by a single lectin for low von Willebrand disease testing and were used to explore the key role of VWF glycan sialylation in regulating its clearance.

[0096] The binding affinity of specific lectins to lectins of healthy and low VWF blood samples was used to map the VWF glycosylation profiles that characterize the surface glycosylation features of VWF in the blood samples, including:

[0097] The lectin binding force corresponding to the specific lectin extracted from the effective sample of the healthy blood sample is plotted to form the VWF glycosylation profile of the healthy blood sample representing the surface glycosylation characteristics of VWF in the healthy blood sample;

[0098] The lectin binding force corresponding to the specific lectin extracted from the effective sample of the low VWF blood sample was plotted to form the VWF glycosylation profile of the low VWF blood sample to characterize the VWF surface glycosylation characteristics of the low VWF blood sample.

[0099] The lectin chip includes Jack Bean agglutinin A, Snowdrop agglutinin, Phaseolus vulgaris agglutinin, Phaseolus vulgaris agglutinin, Interstitial Datura agglutinin, Crystalline crystal agglutinin, Purpura leukohemagglutinin, Amorpha fruticosa agglutinin, Black ginseng agglutinin, Leukohemagglutinin, Euclid agglutinin-I, Lotus root agglutinin, Ricinus communis agglutinin-I, Crystalline Erythrina agglutinin, Peanut agglutinin, Artemisia selengensis agglutinin, Tremella fuciformis agglutinin IB4, Wisteria sinensis agglutinin, Soybean agglutinin, Spiromelia fasciata agglutinin, Amomum villosum agglutinin, Duricula biflora agglutinin, Tremella fuciformis agglutinin II, Wheat germ agglutinin, Tomato agglutinin and Potato agglutinin.

[0100] Specific lectins include black ginseng agglutinin and Ricinus communis agglutinin-I. Black ginseng agglutinin mainly recognizes α2, 6 sialic acid (SA) on N-type sugars, while Ricinus communis agglutinin-I specifically recognizes desialylated β-Gal / β-GalNAc, which can fully analyze the glycosylation characteristics of VWF.

[0101] The present invention incorporates machine vision and utilizes a neural network model for artificial intelligence automated recognition, thereby screening out effective samples of specific VWF glycosylation profiles from all experimental samples, thereby preventing invalid samples from interfering with subsequent analysis of the VWF glycosylation profile, and performing a blood sample control analysis to obtain specific lectins for accurately evaluating VWF glycan characteristics, and subsequently only needs to perform agglutination analysis of black ginseng lectin and ricinus communis lectin-I to obtain VWF glycan characteristics characterizing the blood sample, thereby improving the analysis efficiency and accuracy of plasma VWF surface glycosylation analysis.

[0102] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip, characterized in that: The following steps are involved: Step S1, selecting a group of healthy persons and a group of low VWF patient plasma samples as a group of healthy blood samples and a group of low VWF blood samples respectively, using anti-VWF polyclonal antibody coated ELISA to detect VWF:Ag concentration, and diluting each healthy blood sample and each low VWF blood sample in 0.1% PBS-Tween so that the VWF:Ag concentration in each healthy blood sample and each low VWF blood sample reaches 1 μg / mL; Step S2, incubating each healthy blood sample and each low VWF blood sample with the lectin chip for 1 hour, obtaining a lectin dot matrix fluorescence image of the lectin chip, using the pre-established chip validity judgment model to determine the chip validity of each lectin chip using the lectin dot matrix fluorescence image, and then screening out effective lectin chips from all lectin chips based on the chip validity, and using a chip scanner to read out the lectin binding force of each lectin in the effective lectin chip with the healthy blood sample and the low VWF blood sample after incubating the effective lectin chip for 1 hour; Step S3, using the effective sample recognition model pre-established based on the neural network model to automatically identify the binding force images of each lectin in the effective lectin chip with the lectin binding force of the healthy blood sample and the low VWF blood sample, so as to screen out the effective samples of healthy blood samples and the effective samples of low VWF blood samples for analyzing the specific VWF glycosylation spectrum from a group of healthy blood samples and a group of low VWF blood samples, and perform lectin binding force comparison analysis on the effective samples of healthy blood samples and the effective samples of low VWF blood samples to obtain the specific lectin binding to the low VWF blood sample, and draw the VWF glycosylation spectrum characterizing the surface glycosylation characteristics of VWF in the blood sample based on the binding force of the specific lectin with the lectin of the healthy blood sample and the low VWF blood sample.

2. The method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip according to claim 1, characterized in that: The method of using the lectin dot matrix fluorescence image to determine the chip validity of each lectin chip using a pre-established chip validity determination model includes: The lectin dot matrix fluorescence image of each healthy blood sample is sequentially input into the chip validity determination model, and the chip validity determination model outputs the chip validity of the lectin dot matrix fluorescence image of each healthy blood sample; The lectin dot matrix fluorescence image of each low VWF blood sample is sequentially input into the chip validity determination model, and the chip validity determination model outputs the chip validity of the lectin dot matrix fluorescence image of each low VWF blood sample; A set of defective historical lectin dot matrix fluorescence incomplete images is obtained, and all the historical lectin dot matrix fluorescence incomplete images are manually marked with chip validity category labels, wherein the chip validity category labels include incomplete images and non-incomplete images; Using the historical lectin dot matrix fluorescence defect image as the first input item of the softmax neural network, using the chip validity category label of the historical lectin dot matrix fluorescence defect image as the first output item of the softmax neural network, and performing convolution training on the softmax neural network based on the first input item and the first output item to obtain the chip validity determination model; The model expression of the chip validity determination model is: Label1=softmax(Figure_Vector1); Where Label1 is the expression identifier of the chip validity category label, softmax is the expression identifier of the softmax neural network, and Figure_Vector1 is the expression identifier of the lectin dot matrix fluorescence defect image.

3. The method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip according to claim 2, characterized in that: The method further screens out effective lectin chips from all lectin chips based on chip effectiveness, including: The lectin chips corresponding to the lectin dot matrix fluorescence defective images with chip validity category labels of non-defective in all healthy blood samples are screened out and marked as valid lectin chips for healthy blood samples; The lectin chips corresponding to the lectin dot matrix fluorescence defect images with chip validity category label as non-defective in all low VWF blood samples were screened out and marked as valid lectin chips for low VWF blood samples.

4. The method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip according to claim 3, characterized in that: The method of automatically identifying the binding force images of the binding forces of each lectin in the effective lectin chip with the lectin of the healthy blood sample and the low VWF blood sample by using the effective sample recognition model pre-established based on the neural network model includes: The binding force image of each healthy blood sample is sequentially input into a healthy blood sample valid sample identification model, and the healthy blood sample valid sample identification model outputs the sample validity of each healthy blood sample; sequentially inputting the binding force image of each low VWF blood sample into a low VWF blood sample valid sample identification model, and the low VWF blood sample valid sample identification model outputs the sample validity of each low VWF blood sample; The construction of the effective sample identification model of healthy blood samples includes: Acquire a group of binding force images of historical healthy blood samples, and manually mark the binding force images of all historical healthy blood samples with sample validity category labels, wherein the sample validity category labels include invalid blood samples and valid blood samples; Using the binding force image of the historical healthy blood sample as the second input item of the softmax neural network, using the sample validity category label of the binding force image of the historical healthy blood sample as the second output item of the softmax neural network, and performing convolution training on the softmax neural network based on the second input item and the second output item to obtain the healthy blood sample valid sample recognition model; The model expression of the healthy blood sample effective sample identification model is: Labe2=softmax(Figure_Vector2); Wherein, Labe2 is the expression identifier of the validity category label of the healthy blood sample, softmax is the expression identifier of the softmax neural network, and Figure_Vector2 is the expression identifier of the binding force image of the healthy blood sample; The construction of the effective sample identification model for low VWF blood samples includes: Acquire a group of binding force images of historically low VWF blood samples, and manually mark the binding force images of all historically low VWF blood samples with sample validity category labels, wherein the sample validity category labels include invalid blood samples and valid blood samples; Using the binding force image of the historical low VWF blood sample as the third input item of the softmax neural network, using the sample validity category label of the binding force image of the historical low VWF blood sample as the third output item of the softmax neural network, and performing convolution training on the softmax neural network based on the third input item and the third output item to obtain the low VWF blood sample valid sample recognition model; The model expression of the low VWF blood sample effective sample identification model is: Label3=softmax(Figure_Vector3); Where Label3 is the expression identifier of the low VWF sample validity category label, softmax is the expression identifier of the softmax neural network, and Figure_Vector3 is the expression identifier of the low VWF binding image.

5. The method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip according to claim 4, characterized in that: The screening of the effective samples of healthy blood samples and the effective low VWF blood samples includes: Healthy blood samples with a sample validity category label of valid blood samples are screened out from all healthy blood samples and marked as valid samples of healthy blood samples; The low VWF blood samples with the sample validity category label as valid blood samples among all the low VWF blood samples are screened out and marked as valid samples of low VWF blood samples.

6. The method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip according to claim 5, characterized in that: The obtaining of the specific lectin comprises: n valid samples are selected from both the valid samples of healthy blood samples and the valid samples of low VWF blood samples; The lectin binding force of each lectin in the valid samples of n healthy blood samples is summarized to obtain the lectin binding force vector of each lectin in the healthy blood sample. The function expression of the lectin binding force vector in the healthy blood sample is: X i ={ x ij | i ∈[1,m], j ∈[1,n]}, where X i For healthy blood samples i The lectin binding force vector of each lectin, x ij For healthy blood samples j The valid samples and i The lectin binding force of n lectins, m is the total number of lectins in the effective lectin chip; the lectin binding force of each lectin in the effective samples of n low VWF blood samples is summarized in turn to obtain the lectin binding force vector of each lectin in the low VWF blood sample, and the function expression of the lectin binding force vector in the low VWF blood sample is: Y i ={ y ij | i ∈[1,m], j ∈[1,n]}, where Y i The first i The lectin binding force vector of each lectin, y ij The first j The valid samples and i The lectin binding capacity of each lectin, m is the total number of lectins in the effective lectin chip; Calculate the lectin binding force vector X in healthy blood samples i Binding force vector Y to lectin in low VWF blood samples i and compare the similarity with a similarity threshold, wherein, When the similarity is greater than the similarity threshold, the first i The lectin is a specific lectin; When the similarity is less than or equal to the similarity threshold, the first i The lectin is a non-specific lectin; The similarity is measured using Euclidean distance, and the similarity calculation formula is: Where, P i For X i and Y i similarity, T is the transposition operator.

7. The method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip according to claim 6, characterized in that: The method uses the binding force of the specific lectin to the lectin of the healthy blood sample and the low VWF blood sample to draw a VWF glycosylation profile characterizing the surface glycosylation characteristics of the blood sample VWF, including: The lectin binding force corresponding to the specific lectin extracted from the effective sample of the healthy blood sample is plotted to form the VWF glycosylation profile of the healthy blood sample representing the surface glycosylation characteristics of VWF in the healthy blood sample; The lectin binding force corresponding to the specific lectin extracted from the effective sample of the low VWF blood sample was plotted to form the VWF glycosylation profile of the low VWF blood sample to characterize the VWF surface glycosylation characteristics of the low VWF blood sample.

8. The method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip according to claim 7, characterized in that: The lectin chip includes Jack bean agglutinin A, snowdrop agglutinin, white bean agglutinin, red bean agglutinin, interstitial Datura agglutinin, crystal agglutinin, purpura leukohemagglutinin, Amorpha fruticosa hemagglutinin, black ginseng agglutinin, leukohemagglutinin, Euclid agglutinin-I, lotus root agglutinin, Ricinus communis agglutinin-I, crystal erythrina agglutinin, peanut agglutinin, japanesepine, Tremella fuciformis agglutinin IB4, Wisteria sinensis agglutinin, soybean agglutinin, Helicoverpa oleracea agglutinin, Amomum villosum agglutinin, Duricula biflora agglutinin, Tremella fuciformis agglutinin II, wheat germ agglutinin, tomato agglutinin and potato agglutinin.

9. The method for analyzing the surface glycosylation profile of plasma VWF using a lectin chip according to claim 8, characterized in that: The specific lectins include black ginseng agglutinin and Ricinus communis agglutinin-I.

Images