阅读说明：本技术 一种指导个性化治疗研究的适应性评估标志物筛选算法 (Adaptive evaluation marker screening algorithm for guiding personalized treatment research ) 是由 黄鑫 周洋 苏本哲 于 2020-04-12 设计创作，主要内容包括：本发明提供一种指导个性化治疗研究的适应性评估标志物筛选算法,包括络构建和网络分析,基于相对差异性表达衡量代谢比值特征在治疗有效组样本和治疗无效组样本中的变化构建差异代谢网络,消除了个体差异对于数据分析的影响。同时,基于所构建的差异代谢网络,采用拓扑结构分析方法发现处于通路反应枢纽位置的重要代谢物。提取以所选代谢物为中心的网络子图作为适应性评估标志物,用于在治疗前预测患者是否适合接受某一临床治疗方案,指导精准用药和个性化治疗方案的实施。(The invention provides an adaptive evaluation marker screening algorithm for guiding personalized treatment research, which comprises network construction and network analysis, wherein a differential metabolic network is constructed on the basis of the change of metabolic ratio characteristics measured in a treatment effective group sample and a treatment ineffective group sample in relative differential expression, and the influence of individual difference on data analysis is eliminated. Meanwhile, based on the constructed differential metabolic network, a topological structure analysis method is adopted to discover important metabolites at the position of a pathway reaction junction. Extracting a network subgraph centered on the selected metabolites as an adaptability evaluation marker for predicting whether a patient is suitable for receiving a certain clinical treatment scheme before treatment, and guiding the implementation of precise medication and a personalized treatment scheme.)

1. An adaptive assessment marker screening algorithm for guiding personalized treatment research is characterized by comprising network construction and network analysis;

the network construction comprises the following steps:

1) inputting metabolic data of a treatment effective group and a treatment ineffective group;

2) for any two metabolic features f in the data_i，f_jConstruction of metabolic ratio features r on each type of sample_ij；

3) Calculating each ratio variable r according to equation (1)_ijMean value of S_ij；

Definition of r_ijRepresents metabolite f_iAnd metabolite f_jCharacteristic of the ratio of (S)_ijIs r_ijThe mean over all samples is calculated as in equation (1):

wherein r is_ijtCharacteristic r of metabolic ratio_ijThe expression value on the sample t;

4) according to each metabolic ratio characteristic r_ijMean value of S_ij△ of the metabolic ratio feature in the treatment-effective group and treatment-ineffective group was calculated using formulas (2) and (3)_ijA value;

p_ij(c_l)＝prob(r_ijt<S_ij|y_t＝c_l) (2)

△_ij＝|p_ij(c₁)-p_ij(c₂)|(3)

wherein p is_ij(c_l) Is at c_lEvents in class samples r_ijt<S_ijA priori probability of occurrence △_ij>0 indicates the metabolic ratio characteristic r_ijIn most part c₁The content of the class sample is less than that in c₂Content in class sample and vice versa △_ijThe larger the absolute value is, the characteristic r of the metabolic ratio is shown_ijThe larger the difference between the two types of samples is, the more violent the corresponding metabolic pathway reaction activity is;

5) if △_ijIf the value is larger than or equal to the preset value, two nodes f in the network graph are judged_i，f_jIs indicated by a color, whereas if △_ijThe value is less than or equal to-, then two nodes f in the network graph_i,f_jIs represented in another color.

2. The adaptive assessment marker screening algorithm for guiding personalized therapy research according to claim 1, wherein said network analysis comprises a topology analysis method of a single network map:

in the network diagram, the node with the maximum degree indicates that the metabolite participates in reaction activities of multiple pathways, is positioned at a central hub position, and plays an important biological role in the physiological and pathological changes of a body; extracting a star subgraph formed by the nodes with the maximum degree and the nodes directly connected with the nodes as important network signals; and constructing an adaptability evaluation marker based on the ratio relation between the nodes in the star subgraph, and predicting whether the patient is suitable for receiving a certain clinical treatment scheme before treatment.

3. The adaptive assessment marker screening algorithm for guiding personalized therapy research according to claim 2, wherein said network analysis further comprises a topology analysis method of a plurality of network maps:

when the researched problem is complex, constructing a plurality of network graphs to reflect the essence of the problem; different network diagrams show the metabolic change of an organism under different states, and metabolic signals are divided into two parts, wherein one part of the metabolic signals are used for representing the commonalities of different metabolic mechanisms, the other part of the metabolic signals can be used for describing the differences of the different metabolic mechanisms, and edges with the same or different colors in the different network diagrams are extracted to form important network subgraphs which contain important information closely related to the researched problem; and then, screening key network signals for the extracted network subgraphs by adopting a topological structure analysis mode of a single network graph.

Technical Field

The invention relates to the technical field of biological data analysis, in particular to an adaptive evaluation marker screening algorithm for guiding personalized treatment research.

Background

With the rapid development of system biology, researchers have focused on analyzing the life processes of organisms from the perspective of the integrity of biological systems and sought effective treatment methods for complex diseases. Metabonomics, an important branch of system biology, aims to quantitatively and qualitatively explore dynamic multivariate responses of living organisms to metabolites under the influence of factors such as internal gene mutation, physiological and pathological changes and external environment. The metabolite is the downstream product of the gene and is the final product thereof, the tiny change of the gene expression can be amplified on the metabolite, the embodiment is more obvious, the change of the metabolite concentration can reflect different physiological and pathological states of the organism, and the research on the metabonomics can discover the early metabolic disturbance before the disease symptom changes. Therefore, metabolomics is widely used for research in the fields of clinical disease marker screening and personalized therapy, and has made many important advances.

Diabetes mellitus is a common disease causing disorder of sugar, fat and protein metabolism in the body due to absolute or relative insufficiency of insulin secretion in the body, and has a clinical phenotype that a patient is in a high blood sugar state for a long time. According to the data statistics result of the international diabetes union, the number of diabetic patients in China in 2019 is about 1.164 hundred million, and the method is the country with the most number of diabetic patients. Diabetes is divided into type I diabetes and type II diabetes, with about 95% of patients being type II diabetes. Type II diabetics are often accompanied by symptoms of insulin resistance, complications of the aorta and arteriole vessels, with a probability of cardiovascular disease three times that of the average person. Various complications caused by type II diabetes are the main causes of death and disability of patients, and seriously threaten human health. Sulfonylurea hypoglycemic agents (such as gliclazide, glyburide, etc.) are the common clinical treatment methods for type II diabetes at present. However, the metabolism of the body is influenced by internal heredity, external environmental factors and the like, so that the metabolic phenotype of the body is different, and the individual difference influences the clinical effect of the body on certain drug treatment. Although the gliclazide sustained release preparation is widely applied to clinical treatment of type II diabetes, the ideal treatment effect for part of patients is still not achieved due to the existence of individual difference. The patient population suitable for the gliclazide sustained release preparation is deeply explored, and an adaptability evaluation marker which can be used for guiding the precise medication of the gliclazide sustained release preparation is found, so that a correct personalized treatment scheme is provided for the patients with type II diabetes, and the method has important practical significance.

The living organisms are very complex, and various physiological functions are realized among molecules in the form of a passage or a network to maintain the normal activities of the living organisms, so that close association exists among the molecules. Metabonomics discovers metabolic signals which can be used for representing different physiological and pathological states of organisms by researching the dynamic change of endogenous small molecules in the whole metabolic network or pathway reaction. In metabolomics, the ratio between two metabolites can be interpreted as the chemical reaction by which one metabolite is converted to another metabolite through one or more pathways. Therefore, the difference of the metabolite ratios in the treatment effective patients and the treatment ineffective patients is analyzed, a biological network is constructed based on the change of the metabolite ratios, the important metabolic ratio characteristics with discrimination capability are found, the drug action mechanism can be more comprehensively clarified, and effective adaptive evaluation markers can be screened, so that the clinical research of the personalized drug treatment of the type II diabetes patients is promoted.

In metabonomic studies, the presence of individual variability affects the effectiveness of data analysis. In addition, some data analysis algorithms such as support vector machines and neural networks need to construct complex decision boundaries, which are difficult to be interpreted from a biological perspective, and often limit further clinical applications. Therefore, the data analysis method should be intuitive, efficient, and highly biologically interpretable. The invention considers the importance of metabolic pathway reaction in the life activity of an organism, constructs a biological network based on the change of the metabolic ratio characteristic expression value, describes the difference of the metabolic pathway reaction in different physiological and pathological states of the organism by utilizing a network topological structure, and can intuitively and effectively discover important metabolic signals reflecting the essence of problems. The invention constructs a differential metabolic network by using the change of the metabolic ratio characteristic measured by relative differential expression in the treatment effective group sample and the treatment ineffective group sample, and eliminates the influence of individual difference on data analysis. Meanwhile, based on the constructed differential metabolic network, a topological structure analysis method is adopted to discover important metabolites at the position of a pathway reaction junction. And extracting a network subgraph centered on the selected metabolites as an adaptability evaluation marker for pre-evaluating the clinical treatment effect of a certain treatment scheme on a certain patient, and guiding the implementation of accurate medication and a personalized treatment scheme.

Disclosure of Invention

In order to solve the technical problems brought forward by the background technology, the invention provides an adaptive evaluation marker screening algorithm for guiding personalized treatment research, which takes effective and ineffective clinical sample data under a certain treatment scheme as a research object, takes searching of an adaptive evaluation marker as a research target, takes construction and analysis of a metabolic network as a research means, and explores real responses of different patients to the certain treatment scheme, thereby being beneficial to promoting research and application of clinical personalized treatment.

In order to achieve the purpose, the invention adopts the following technical scheme:

an adaptive assessment marker screening algorithm for guiding personalized treatment research comprises network construction and network analysis;

the network construction comprises the following steps:

1) inputting metabolic data of a treatment effective group and a treatment ineffective group;

2) for any two metabolic features f in the data_i，f_jConstruction of metabolic ratio features r on each type of sample_ij；

3) Calculating each ratio variable r according to equation (1)_ijMean value of S_ij；

Definition of r_ijRepresents metabolite f_iAnd metabolite f_jCharacteristic of the ratio of (S)_ijIs r_ijThe mean over all samples is calculated as in equation (1):

wherein r is_ijtCharacteristic r of metabolic ratio_ijThe expression value on the sample t;

4) according to each metabolic ratio characteristic r_ijMean value of S_ij△ of the metabolic ratio feature in the treatment-effective group and treatment-ineffective group was calculated using formulas (2) and (3)_ijA value;

p_ij(c_l)＝prob(r_ijt<S_ij|y_t＝c_l) (2)

△_ij＝|p_ij(c₁)-p_ij(c₂)| (3)

wherein p is_ij(c_l) Is at c_lEvents in class samples r_ijt<S_ijA priori probability of occurrence △_ij>0 indicates the metabolic ratio characteristic r_ijIn most part c₁The content of the class sample is less than that in c₂Class of contents in samples and vice versa △_ijThe larger the absolute value is, the characteristic r of the metabolic ratio is shown_ijThe larger the difference between the two types of samples is, the more violent the corresponding metabolic pathway reaction activity is;

5) if △_ijIf the value is larger than or equal to the preset value, two nodes f in the network graph are judged_i，f_jIs indicated by a color, whereas if △_ijThe value is less than or equal to-, then two nodes f in the network graph_i,f_jIs represented in another color.

The metabolic network constructed by the invention can be expressed as follows: definitions G ═ (v (G), e (G), w (G)) are undirected networks with weights, v (G) ═ F denote node sets, e (G) { (F)_i,f_j)|f_i,f_j∈F,△_ij≥}U{(f_i,f_j)|f_i,f_j∈F,△_ij≦ represents the edge set. W (g) represents a weight set of edges, and is defined as w (g) ═ w (f)_i,f_j)＝△_ij|(f_i,f_j) ∈ E (G) }. if w (f)_i,f_j) If not, node f_iAnd node f_jThe connecting edges between the two are expressed by a color; if w (f)_i,f_j) Less than or equal to the node f_iAnd node f_jThe connecting edges between are represented in another color.

(II) the network analysis comprises a topology analysis method of a single network diagram and a topology analysis method of a plurality of network diagrams;

1) the topological structure analysis method of the single network graph comprises the following steps:

in the network diagram, the node with the maximum degree indicates that the metabolite participates in reaction activities of multiple pathways, is positioned at a central hub position, and plays an important biological role in the physiological and pathological changes of a body; extracting a star subgraph formed by the nodes with the maximum degree and the nodes directly connected with the nodes as important network signals; constructing an adaptability evaluation marker based on the ratio relation between nodes in the star subgraph, and predicting whether a patient is suitable for receiving a certain clinical treatment scheme before treatment;

2) the topological structure analysis method of a plurality of network graphs comprises the following steps:

when the researched problem is complex, constructing a plurality of network graphs to reflect the essence of the problem; different network diagrams show the metabolic change of an organism under different states, and metabolic signals are divided into two parts, wherein one part of the metabolic signals are used for representing the commonalities of different metabolic mechanisms, the other part of the metabolic signals can be used for describing the differences of the different metabolic mechanisms, and edges with the same or different colors in the different network diagrams are extracted to form important network subgraphs which contain important information closely related to the researched problem; and then, screening key network signals for the extracted network subgraphs by adopting a topological structure analysis mode of a single network graph.

Compared with the prior art, the invention has the beneficial effects that:

1) the invention constructs a differential metabolic network by using the change of the metabolic ratio characteristic measured by relative differential expression in the treatment effective group sample and the treatment ineffective group sample, and eliminates the influence of individual difference on data analysis. Meanwhile, based on the constructed differential metabolic network, a topological structure analysis method is adopted to discover important metabolites at the position of a pathway reaction junction. And extracting a network subgraph centered on the selected metabolites as an adaptability evaluation marker for pre-evaluating the clinical treatment effect of a certain treatment scheme on a certain patient, and guiding the implementation of accurate medication and a personalized treatment scheme.

2) Due to the existence of individual differences, accurate medical treatment can enable patients to receive correct treatment schemes as soon as possible, and clinical treatment effects are improved. Therefore, prior to taking a certain treatment regimen, a pre-evaluation test of the efficacy of the treatment is required on the patient. The project measures the change of metabolic ratio characteristics in treatment effective group and treatment ineffective group samples under a certain scheme based on a relative difference expression method, constructs a difference metabolic network, and systematically describes the difference of metabolic pathway reaction activities in different physiological and pathological states in a network mode. Based on the constructed differential metabolic network, a topological structure analysis method is adopted to find important metabolites at the position of a pathway reaction active junction, and a star subgraph taking the metabolites as the center is extracted. And constructing an adaptability evaluation marker set by using the association relation among the metabolites in the extracted star subgraph, thereby predicting whether a certain treatment scheme is effective for the rehabilitation of a certain patient.

Detailed Description

The following describes in detail specific embodiments of the present invention.

The metabolites are related and interacted, and when the organism shows physiological and pathological changes, the disturbance occurs not only in one metabolic molecule but also in the whole module or metabolic network. The invention provides a new difference metabolic network construction algorithm DMNC, which deeply explores the difference of pathway reaction activities in a treatment effective patient and a treatment ineffective patient from the network perspective and finds an adaptability evaluation marker which can be used for guiding personalized treatment. Changing F to F₁,f₂,…,f_mDefining a feature set, wherein m represents the number of features; x ═ X₁,x₂,...,x_nDefining a sample set, wherein n represents the number of samples; c ═ C₁,c₂Is defined as a set of class labels, c₁Represents a therapeutically effective group, c₂Indicating a treatment-ineffective group; y ═ Y₁,y₂,…,y_nIs defined as a vector of n sample indices, y_i∈ C denotes sample x_iClass label of (2).

Definition of r_ijRepresents metabolite f_iAnd metabolite f_jCharacteristic of the ratio of (S)_ijIs r_ijThe mean over all samples is calculated as in equation (1):

wherein r is_ijtCharacteristic r of metabolic ratio_ijExpression value on sample t. Relative differential expression metabolic ratio characteristic r is measured by the following formula_ijChanges in expression values in the treatment-effective and treatment-ineffective samples:

p_ij(c_l)＝prob(r_ijt<S_ij|y_t＝c_l) (2)

△_ij＝|p_ij(c₁)-p_ij(c₂)| (3)

wherein p is_ij(c_l) Is at c_lEvents in class samples r_ijt<S_ijA priori probability of occurrence △_ij>0 indicates the metabolic ratio characteristic r_ijIn most part c₁The content of the class sample is less than that in c₂Class of contents in samples and vice versa △_ijThe larger the absolute value is, the characteristic r of the metabolic ratio is shown_ijThe greater the difference between the two types of samples, the more vigorous the corresponding metabolic pathway response.

The metabolic network constructed by the invention can be expressed as follows: definitions G ═ (v (G), e (G), w (G)) are undirected networks with weights, v (G) ═ F denote node sets, e (G) { (F)_i,f_j)|f_i,f_j∈F,△_ij≥}U{(f_i,f_j)|f_i,f_j∈F,△_ij≦ represents the edge set. W (g) represents a weight set of edges, and is defined as w (g) ═ w (f)_i,f_j)＝△_ij|(f_i,f_j) ∈ E (G) }. if w (f)_i,f_j) If not, node f_iAnd node f_jThe connecting edges between the two are expressed by a color; if w (f)_i,f_j) Less than or equal to the node f_iAnd node f_jThe connecting edges between are represented in another color. The differential metabolism network constructed based on the method can be used for screening adaptability evaluation markers and providing help for individualized treatment of patients, and the method is suitable for the fields of disease omics data analysis, transformation medicine and the like.

The technical scheme adopted by the invention is as follows:

network construction

Step 1: inputting metabolic data of a treatment effective group and a treatment ineffective group;

step 2: for any two metabolic features f in the data_i，f_jConstruction of metabolic ratio features r on each type of sample_ij；

And step 3: calculating each ratio feature r according to equation (1)_ijMean value of S_ij；

And 4, step 4: according to the metabolic ratio characteristic r_ijMean value of S_ijThe difference △ between the treatment effective group and the treatment ineffective group is calculated using the formulas (2) and (3)_ij；

Step 5, if △_ijIf the value is larger than or equal to the preset value, two nodes f in the network graph are judged_i，f_jIs indicated by a red edge, otherwise, if △_ijThe value is less than or equal to-, then two nodes f in the network graph_i，f_jThe continuous side of (a) is indicated by a green side.

(II) network analysis

The network constructed by the invention can represent the difference of metabolic pathway reaction activities of organisms in different states. And (3) discovering important network metabolic signals closely related to the researched problem by adopting a topological structure analysis method for the differential metabolic network.

1) Topological structure analysis method of single network graph

In the network diagram, the node with the largest degree indicates that the corresponding metabolite participates in the reaction activities of multiple pathways, is positioned at the central hub position, and plays an important biological role in the physiological and pathological changes of the organism. Therefore, the invention extracts the node with the maximum degree and the star subgraph formed by the nodes directly connected with the node as important network signals. And constructing an adaptability evaluation marker based on the ratio relation between the nodes in the star subgraph, and predicting whether the patient is suitable for receiving a certain clinical treatment scheme before treatment.

2) Topological structure analysis method of multiple network graphs

When the problem to be researched is complex, a plurality of network graphs are required to be constructed to reflect the nature of the problem. Different network diagrams show the metabolic changes of the organism in different states, some metabolic signals can be used for characterizing the commonality of different metabolic mechanisms, and some metabolic signals can be used for describing the difference of different metabolic mechanisms. Aiming at a specific research problem, the invention extracts the edges with the same (or different) colors in different network graphs to form an important network subgraph which contains important information closely related to the research problem. And then, screening key network signals for the extracted network subgraphs by adopting a topological structure analysis mode of a single network graph.