Heterogeneous network fusion and importance-based lncRNA function enrichment analysis method
By constructing the lncRNA-lncRNA functional similarity network and calculating the importance score, combining disease, miRNA and expression profile associations, the problem of inaccurate prediction of lncRNA functional differences and unconsidered in traditional methods is solved, and lncRNA functional enrichment analysis is realized Improves comprehensiveness and accuracy.
Patent Information
- Application Number
- CN202411310559.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-20
- Publication Date
- 2024-12-20
AI Technical Summary
The traditional lncRNA functional enrichment analysis method cannot comprehensively and accurately predict the regulatory function of lncRNAs, and ignores the size of different lncRNAs playing a role in biological processes, resulting in incomplete enrichment analysis results, and treating each lncRNA equally, and failing to Consider fully the differences in importance under the same function.
Using a method based on heterogeneous network fusion and importance, the lncRNA-lncRNA functional similarity network was constructed, and the importance score of lncRNA was calculated, and the correlation between disease, miRNA and expression profile was combined to perform functional enrichment analysis and amplify lncRNA. List to improve the comprehensiveness and accuracy of the analysis.
The comprehensiveness and accuracy of lncRNA functional enrichment analysis has been achieved, and the potential set of annotations can be better identified, reflect the degree of influence of different lncRNAs under the same function, providing an important reference value for the lncRNA regulatory mechanism.
Smart Images
Figure CN119170108A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of lncRNA functional enrichment analysis, and specifically is a lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance. Background Art
[0002] Most of the DNA in the human genome can be transcribed into RNA, but only about 2% of RNA can encode proteins, and the rest are non-coding RNA (ncRNA). As research continues to deepen, more and more evidence shows that non-coding RNA has an important regulatory role. Non-coding RNA with a length greater than 200 nt is called long non-coding RNA (lncRNA). It was originally considered to be a byproduct of gene transcription and does not have the function of regulating biological organisms. However, a large number of studies have confirmed that lncRNA plays an important regulatory role in biological processes such as gene expression regulation, chromatin structure maintenance, and transcription factor regulation. In addition, lncRNA has the characteristics of clustering into groups and exerting regulatory effects in the form of a collection during the regulatory process, which has transformed the research on the regulatory function of lncRNA from the analysis of the function of a single lncRNA to the analysis of the function of lncRNA collections at the system level, resulting in functional enrichment analysis becoming one of the important methods for studying the regulatory function of lncRNA in the era of big data.
[0003] Functional enrichment analysis is a method for explaining whether a gene set is significantly enriched in a certain function, thereby judging the biological significance of the gene list. Different lncRNAs play different regulatory functions in different biological processes, but the traditional lncRNA regulatory function network only describes the regulatory function of lncRNA from one aspect, resulting in the inability to fully and accurately predict the regulatory function of lncRNA, and thus the inability to effectively amplify the lncRNA list, making the enrichment analysis results incomplete. In addition, the traditional lncRNA enrichment analysis method treats each lncRNA equally, assuming that all lncRNAs play the same role under the same function, which is inconsistent with the actual biological background. In fact, the effects of abnormalities of different lncRNAs on the body vary greatly, and the importance of each lncRNA in the same biological function or pathway also varies. Therefore, in enrichment analysis, the importance of lncRNA itself needs to be fully considered. Based on this, the present invention proposes a lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance. Summary of the invention
[0004] In view of the deficiencies of the prior art, the technical problem to be solved by the present invention is to propose a lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance.
[0005] The present invention solves the technical problem by adopting the following technical solution:
[0006] A lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance, characterized by comprising the following steps:
[0007] S1. Obtain lncRNA functional annotation set;
[0008] S2. Construct lncRNA functional similarity matrices based on lncRNA-disease association, lncRNA-miRNA association and expression profile, respectively, and construct lncRNA feature vectors based on these matrices; calculate the Pearson correlation coefficient between any two lncRNA feature vectors, use lncRNA as a node and the Pearson correlation coefficient as the weight of the edge, and construct a lncRNA-lncRNA functional similarity network; amplify the list of lncRNAs to be analyzed based on the lncRNA-lncRNA functional similarity network;
[0009] S3. Select lncRNA importance attributes for the biological process of enrichment analysis, standardize the importance attribute values of each lncRNA, and obtain the importance score of each lncRNA;
[0010] S4. Calculate the sum of the importance scores of all lncRNAs in the corresponding set according to formula (10), and calculate the distribution probability P of the lncRNA list on the lncRNA functional annotation set according to formula (11);
[0011]
[0012] Where N′ is the sum of the importance scores of all lncRNAs in the background lncRNA set N covered by the functional enrichment analysis, n′ is the sum of the importance scores of all lncRNAs in the lncRNA functional annotation set n, M′ is the sum of the importance scores of all lncRNAs in the lncRNA list M, k′ is the sum of the importance scores of all lncRNAs in the intersection of the lncRNA list and the lncRNA functional annotation set, is the importance score of the i-th lncRNA, and Q is the regulatory factor;
[0013] If the distribution probability is less than or equal to the set threshold, it indicates that there is significant enrichment between the lncRNA list and the lncRNA functional annotation set, otherwise there is no significant enrichment.
[0014] Furthermore, the process of constructing the lncRNA functional similarity matrix based on lncRNA-disease association is as follows:
[0015] Define DG(u) and DG(v) as the associated disease sets associated with lncRNA u and v collected from the lncRNA-disease association data, and calculate the disease coefficient S(dG) associated with lncRNA u in the associated disease set DG(v) according to the following formula: u , DG(v)) and the disease coefficient S(d v ,DG(u));
[0016]
[0017] Where, d u ,d v denote the diseases associated with lncRNA u and v, respectively, and DSS(·) represents the semantic similarity between diseases;
[0018] The similarity S of the associated disease set DG(u) relative to DG(v) is calculated according to the following formula: u→v And the similarity S of the associated disease set DG(v) relative to DG(u) v→u :
[0019]
[0020] The functional similarity FS(u,v) of lncRNA u and v based on lncRNA-disease association was calculated according to the following formula:
[0021]
[0022] In the formula, |·| represents the total number of diseases in the associated disease set;
[0023] According to the above process, the functional similarity of any two lncRNAs based on the lncRNA-disease association is calculated, and the functional similarity is used as a matrix element to construct a lncRNA functional similarity matrix based on the lncRNA-disease association.
[0024] Furthermore, the semantic similarity calculation process between diseases is as follows:
[0025] The lncRNA associated with disease A and the corresponding terms in the medical disease vocabulary form a directed acyclic graph DAG A , DAG A The semantic contribution value of each term in to disease A is calculated by the following formula:
[0026]
[0027] In the formula, is the semantic contribution of term t to disease A, Δ is the weight of the edge connecting term t and t′, is the semantic contribution of term t′ to disease A, C(t) is the subset of term t, P t is the specific contribution factor of term t;
[0028] The semantic value SV(A) of disease A is calculated by the following formula:
[0029]
[0030] Where, T A is the term set of disease A;
[0031] Based on the intersection of the term sets of diseases A and B, the semantic similarity DSS(A,B) between diseases A and B is calculated;
[0032]
[0033] In the formula, is the semantic contribution of term t to disease B, SV(B) is the semantic value of disease B, T B is the term set of disease B.
[0034] Furthermore, the process of constructing the lncRNA functional similarity matrix based on lncRNA-miRNA association is as follows:
[0035] The miRNAs related to lncRNA u and v are listed in miRNA_list u and miRNA_list v , take the intersection of the two lists to get the list miRNA_list; construct a multidimensional vector based on the list miRNA_list to get the first miRNA associated feature vector Vector of lncRNA u and v u_1 and Vector v_1 ;
[0036] miRNA_list u The set K of all miRNAs in u , list miRNA_list v The set K of all miRNAs in v , find the set K from the miRNA function similarity matrix u For K v The similarity between all miRNAs in lncRNA u is obtained, and the second miRNA associated feature vector Vector is obtained. u_2 ; Similarly, the second miRNA associated feature vector Vector of lncRNA v is obtained v_2 ;
[0037] The first miRNA associated feature vector of lncRNA u is Vector u_1 Vector associated with the second miRNA u_2 Connect the two ends to get the miRNA-associated feature vector Vector of lncRNA u u ; Similarly, we get the miRNA-associated feature vector Vector of lncRNA v v ;
[0038] According to formula (8), the two miRNA association feature vectors Vector u and Vector v The cosine similarity Sim(lncRNA u,lncRNA v) is used as the functional similarity of lncRNA u and v based on lncRNA-miRNA association;
[0039]
[0040] The above process was repeated to calculate the functional similarity of any two lncRNAs based on lncRNA-miRNA association, and the lncRNA functional similarity matrix based on lncRNA-miRNA association was constructed.
[0041] Furthermore, the process of constructing the lncRNA functional similarity matrix based on expression profile is as follows: construct an expression profile feature vector for each lncRNA, calculate the Pearson correlation coefficient between the expression profile feature vectors of two lncRNAs, use the coefficient as the functional similarity of the two lncRNAs based on expression profile, and construct the lncRNA functional similarity matrix based on expression profile.
[0042] Furthermore, the process of constructing the lncRNA feature vector is as follows: for the lncRNA that only appears in one of the functional similarity matrices, the row vector of the lncRNA is extracted from the corresponding functional similarity matrix as the feature vector; for the lncRNA that appears in two or three functional similarity matrices at the same time, the row vector of the lncRNA is extracted from the corresponding functional similarity matrix respectively, and the three row vectors are concatenated to obtain the feature vector of the lncRNA.
[0043] Furthermore, the amplification process of the lncRNA list to be analyzed is as follows: setting the similarity coefficient, taking all lncRNAs in the lncRNA list to be analyzed as seed nodes, discarding the edges with weights less than the similarity coefficient in the lncRNA-lncRNA functional similarity network, and amplifying by random walk.
[0044] Compared with the prior art, the present invention has the following beneficial effects:
[0045] 1. The present invention proposes a novel method for amplifying the lncRNA list. Taking into full consideration the relationship between lncRNA and disease, miRNA and expression profile, lncRNA functional similarity matrices based on lncRNA-disease association, lncRNA-miRNA association and expression profile are constructed respectively, and the feature vector of lncRNA is extracted from the functional similarity matrix by using the network representation learning method, and the Pearson correlation coefficient of two lncRNA feature vectors is calculated, which is used as the weight of the edge connecting the two lncRNA nodes, and then the lncRNA-lncRNA functional similarity network is constructed. Based on the network, the user-input lncRNA list is amplified by random walk, which helps to discover some potential functional annotation sets and improve the comprehensiveness and accuracy of functional enrichment analysis.
[0046] 2. In the research on lncRNA function prediction and analysis in the prior art, the size of the role played by different lncRNAs in biological processes is ignored, and the role of all lncRNAs is regarded as the same, and all lncRNAs are treated equally. However, the present invention proposes a lncRNA importance score for the mutual relationship between lncRNA and other biological molecules and its regulatory mechanism in biological processes, quantifies the importance attribute value of lncRNA, and different lncRNAs have different importance scores, which is of great significance for revealing the lncRNA regulatory mechanism. The lncRNA importance score can intuitively reflect the degree of influence of different lncRNAs under the same function, helping researchers to quickly determine the key lncRNAs in the functional annotation set, and providing important reference value for the subsequent research on lncRNA regulatory mechanisms and the treatment of related diseases. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION
[0048] Specific embodiments are given below in conjunction with the accompanying drawings. The specific embodiments are only used to introduce the technical solutions of the present invention in detail and are not intended to limit the protection scope of the present application.
[0049] The present invention provides a lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance (hereinafter referred to as method, see Figure 1 ), comprising the following steps:
[0050] S1. Collect lncRNA related data and obtain lncRNA functional annotation set;
[0051] Comprehensive collection of lncRNA functional annotation sets is the basis for lncRNA set enrichment analysis. lncRNA functional annotation information is obtained from multiple public datasets. 33 types of lncRNA functional annotation sets are obtained through sorting. The same functional annotation sets under the same type are merged, and finally a functional annotation set containing 342,857 lncRNAs is obtained.
[0052] S2, constructing a lncRNA-lncRNA functional similarity network, and effectively amplifying the list of lncRNAs to be analyzed based on the network;
[0053] (1) Construction of lncRNA functional similarity matrix based on lncRNA-disease association
[0054] The semantic similarity between diseases was calculated using the directed acyclic graph (DAG) of disease terms corresponding to known disease-associated lncRNAs and the Medical Disease Thesaurus (MeSH). [1] For disease A, its DAG is expressed as DAG A ={T A ,E A}, T A is the term set of disease A, E A It is DAG A The set of edges; each exists in the DAG A The term t∈T in A There is a semantic contribution value for disease A, which is calculated by the following formula:
[0055]
[0056] In the formula, is the semantic contribution of term t to disease A, Δ is the weight of the edge connecting term t and t′, is the semantic contribution value of term t′ to disease A, C(t) is the subset of term t; P t is the specific contribution factor of term t, whose value varies with different versions of medical disease vocabulary and is defined as follows:
[0057]
[0058] Where R represents the set of all terms in the medical disease vocabulary, r represents a term in the medical disease vocabulary, D represents the total number of diseases in the medical disease vocabulary, Dags(r) represents the number of directed acyclic graphs containing term r, and Dags(t) represents the number of directed acyclic graphs containing term t.
[0059] The semantic value SV(A) of disease A is DAG A The sum of the semantic contribution values of all terms in disease A is calculated as follows:
[0060]
[0061] For the directed acyclic graph of two diseases, if there are more identical terms, the number of identical nodes in the directed acyclic graph is greater, and the identical nodes are closer to the target node, it means that the two diseases have higher semantic similarity. Therefore, based on the intersection of the term set of disease A and the term set of disease B, the semantic similarity DSS(A,B) between disease A and disease B is calculated, and the calculation formula is as follows:
[0062]
[0063] In the formula, is the semantic contribution of term t to disease B, SV(B) is the semantic value of disease B, T B is the term set of disease B;
[0064] DG(u) and DG(v) are defined as the associated disease sets associated with lncRNA u and v collected from the lncRNA-disease association data, respectively. The functional similarity between lncRNAu and v is calculated based on the semantic similarity between the diseases in the associated disease sets DG(u) and DG(v). First, the disease coefficient S(d u , DG(v)) and the disease coefficient S(d v ,DG(u)), the calculation formula is as follows:
[0065]
[0066] Where, d u ,d v They represent diseases associated with lncRNA u and v, respectively;
[0067] Then, all the disease coefficients related to lncRNA u in the associated disease set DG(v) are accumulated to obtain the similarity S of the associated disease set DG(u) relative to DG(v) u→v ; Similarly, we get the similarity S of the associated disease set DG(v) relative to DG(u) v→u ; The calculation formula is as follows:
[0068]
[0069] Finally, the functional similarity FS(u,v) between lncRNA u and v based on lncRNA-disease association is calculated. The calculation formula is as follows:
[0070]
[0071] In the formula, |·| represents the total number of diseases in the associated disease set;
[0072] The above process was repeated to calculate the functional similarity of any two lncRNAs based on the lncRNA-disease association, and the functional similarity was used as the matrix element to construct the lncRNA functional similarity matrix based on the lncRNA-disease association.
[0073] (2) Construction of lncRNA functional similarity matrix based on lncRNA-miRNA association
[0074] The miRNAs related to lncRNA u and v are listed in miRNA_list u and miRNA_list v , the two lists are intersected to obtain the list miRNA_list, which covers all miRNAs that interact with lncRNA u and v, reflecting the situation of common target miRNAs between lncRNA u and v [2] . Construct an n-dimensional vector based on the list miRNA_list, where each dimension corresponds to a miRNA in the list miRNA_list. If a miRNA appears in the list miRNA_list u If the corresponding dimension is 1, then it is set to 0; after traversing all miRNAs in the list miRNA_list, the first miRNA associated feature vector Vector of lncRNA u is obtained u_1 Similarly, construct the first miRNA associated feature vector Vector of lncRNA v v_1 .
[0075] miRNA_list u The set K of all miRNAs in u , list miRNA_list v The set K of all miRNAs in v Find the set K from the miRNA function similarity matrix u For K v The similarity between all miRNAs Sim(K u ,K v ), and obtain the second miRNA associated feature vector Vector of lncRNA u u_2 ; Similarly, the second miRNA associated feature vector Vector of lncRNA v is obtained v_2 ; The first miRNA associated feature vector Vector of lncRNA u u_1Vector associated with the second miRNA u_2 Connect the two ends to get the miRNA-associated feature vector Vector of lncRNA u u Similarly, the first miRNA associated feature vector Vector of lncRNA v v_1 Vector associated with the second miRNA v_2 Connect the two ends to get the miRNA-associated feature vector Vector of lncRNA v v ; Calculate the two miRNA association feature vectors according to formula (8) u and Vector v The cosine similarity Sim(lncRNA v,lncRNA v) is taken as the functional similarity of lncRNA u and v based on lncRNA-miRNA association;
[0076]
[0077] The above process was repeated to calculate the functional similarity of any two lncRNAs based on lncRNA-miRNA association, and the lncRNA functional similarity matrix based on lncRNA-miRNA association was constructed.
[0078] (3) Construction of lncRNA functional similarity matrix based on expression profile
[0079] A large number of studies have shown that lncRNAs with similar regulatory functions often have similar expression profiles. Obtain lncRNA expression profiles of 24 human organs or tissues [3] ; Construct an expression profile feature vector for each lncRNA, and the value of each dimension is the expression level of the lncRNA in a certain organ or tissue; Calculate the Pearson correlation coefficient between the expression profile feature vectors of two lncRNAs, and use it as the functional similarity of the two lncRNAs based on the expression profile, so as to construct the lncRNA functional similarity matrix based on the expression profile.
[0080] (4) Construct lncRNA feature vectors and construct lncRNA-lncRNA functional similarity network based on lncRNA feature vectors
[0081] For lncRNAs that only appear in one of the functional similarity matrices, the row vector of the lncRNA is extracted from the corresponding functional similarity matrix as the final feature vector; for lncRNAs that appear in two or three functional similarity matrices at the same time, the row vector of the lncRNA is extracted from the corresponding functional similarity matrix respectively, and the three row vectors are concatenated to obtain the feature vector of the lncRNA. The Pearson correlation coefficient between the feature vectors of two lncRNAs is calculated to reflect the closeness between the lncRNAs; lncRNAs are used as nodes, and the Pearson correlation coefficient between the feature vectors of lncRNAs is used as the weight of the edge to construct the lncRNA-lncRNA functional similarity network.
[0082] (5) Amplify the lncRNA list to be analyzed
[0083] A similarity coefficient reflecting the degree of amplification was set, all lncRNAs in the lncRNA list to be analyzed were used as seed nodes, the edges with weights less than the similarity coefficient in the lncRNA-lncRNA functional similarity network were discarded, and the lncRNA list to be analyzed was amplified using a random walk method to obtain the amplified lncRNA list.
[0084] S3. Select lncRNA importance attributes for the biological process of enrichment analysis, standardize the importance attribute values of each lncRNA, and obtain the importance score of each lncRNA;
[0085] The lncRNA importance attribute is an important factor that affects the regulatory function of lncRNA, and the importance score is the quantification of the attribute importance. For example, in the biological process of cell differentiation, the higher the expression level of a certain lncRNA, the more important the role played by the lncRNA in the biological process. At this time, the expression level is an indicator of the importance of lncRNA. The more transcription factors regulated by lncRNA, the more biological processes the lncRNA may be involved in. Therefore, the number of transcription factors is also an indicator of the importance of lncRNA. Each lncRNA plays a unique and critical role in the development, growth and function of an organism. The importance of different lncRNAs in different biological processes varies, and the measurement of lncRNA importance is also diverse. By consulting a large amount of literature and in-depth analysis of various biological molecules that interact with lncRNA, this embodiment selects 6 types of lncRNA importance attributes, namely lncRNA gene importance (GIC) [4], lncRNA expression level, number of lncRNA upstream regulatory factors, number of lncRNA downstream regulatory factors, number of lncRNA-disease associations, and number of lncRNA-enhancer associations. Among them, the lncRNA gene importance value is derived from the lncRNA sequence information, and the lncRNA expression level value is derived from LncRBase 2.0 [5] Database, the number of lncRNA upstream regulatory factors comes from CHIPbase 3.0 [6] 、LncRNome [7] and LncSEA 2.0 [8] Database, the number of lncRNA downstream regulatory factors comes from LncRNA2Target 2.0 [9] and LncSEA 2.0 database, the number of lncRNA-disease associations and the number of lncRNA-enhancer associations are derived from the lncRNA functional annotation set collected in this example.
[0086] Different biological processes select different lncRNA importance attributes, so for different biological processes, the most relevant lncRNA importance attributes are selected for functional enrichment analysis. For any type of lncRNA importance attribute, the importance attribute values of different lncRNAs are different, so the importance attribute values of each lncRNA are standardized to obtain the importance score of each lncRNA. At the same time, in order to avoid the importance score of some lncRNAs being too high, the regulatory factor Q is introduced as an index to adjust the lncRNA importance score. The calculation formula is as follows:
[0087]
[0088] In the formula, is the importance score of the ith lncRNA, w avg It is the mean of all lncRNA importance attribute values of a certain class after range standardization. is the importance attribute value of the i-th lncRNA, max(w) and min(w) are the maximum and minimum lncRNA importance attribute values of a certain class, respectively;
[0089] After the above standardization, the final six categories of lncRNA importance scores with an average of 1 were obtained. The average lncRNA importance score in the traditional enrichment analysis method is also 1, which makes the lncRNA enrichment analysis results of the fusion importance score comparable with the results of the traditional enrichment analysis method. Prioritizing lncRNAs according to importance scores can more accurately predict the regulatory effects of different lncRNAs in specific biological processes. The higher the importance score, the greater the regulatory effect.
[0090] S4. lncRNA functional enrichment analysis based on importance scores;
[0091] For the functional enrichment analysis of lncRNAs with fusion importance, not only the number of lncRNAs in the intersection of the amplified lncRNA list and the given lncRNA functional annotation set should be considered, but also the importance of each lncRNA. By calculating the weighted sum of the importance scores of all lncRNAs in the corresponding set, the enrichment degree of the lncRNA list in a specific functional pathway and the importance of each lncRNA in the enrichment analysis are comprehensively evaluated. The calculation formula is as follows:
[0092]
[0093] Wherein, N′ is the sum of the importance scores of all lncRNAs in the background lncRNA set N covered by the functional enrichment analysis, n′ is the sum of the importance scores of all lncRNAs in the given lncRNA functional annotation set n, M′ is the sum of the importance scores of all lncRNAs in the amplified lncRNA list M, and k′ is the sum of the importance scores of all lncRNAs in the intersection of the amplified lncRNA list and the given lncRNA functional annotation set;
[0094] According to the hypergeometric distribution probability calculation formula of formula (11), the distribution probability of the amplified lncRNA list on the given lncRNA functional annotation set is calculated;
[0095]
[0096] Where, P is the distribution probability of the amplified lncRNA list on a given lncRNA functional annotation set;
[0097] If the distribution probability is less than or equal to the set threshold (0.05 is selected in this embodiment), it indicates that the amplified lncRNA list is significantly enriched with the given lncRNA function annotation set, otherwise there is no significant enrichment, and the lncRNA function enrichment analysis is completed.
[0098] Example
[0099] In this example, the lncRNA differential expression list of breast cancer is selected as the lncRNA list to be analyzed. The list has log2FC>1, P adj <0.05 were obtained from TCGA and CirncRNAnet databases. The method of the present invention and two existing methods (lncSEA2.0 and TLSEA) were used respectively.
[10] ) The top 10 enrichment analysis results under the same lncRNA functional annotation set (disease) are shown in Tables 1 and 2. In the experiment, the expression level of lncRNA in breast tissue was selected as the lncRNA importance score, and the value of the regulatory factor Q was 3.
[0100] Table 1 Top 10 enrichment analysis results of this method and TLSEA (P value)
[0101]
[0102]
[0103] Table 2 Top 10 enrichment analysis results of this method and LncSEA2.0 (P value)
[0104]
[0105] As can be seen from Tables 1 and 2, the P values of this patented method are smaller than those of the lncSEA2.0 and TLSEA methods. When selecting different thresholds to statistically analyze the significance of the enrichment analysis results, researchers can obtain more enrichment analysis results, broadening their thinking and vision.
[0106] The experimental results also show the importance scores of each lncRNA hit in the breast cancer annotation set, see Table 3.
[0107] Table 3 Importance scores of each lncRNA hit in the breast cancer annotation set
[0108]
[0109] Table 3 intuitively shows the importance scores of different lncRNAs, which can help to quickly identify key lncRNAs and improve the efficiency of enrichment analysis. In the experiment, it was found that due to the introduction of the importance score of lncRNA by the method of the present invention, some functional annotation sets with a small number of overlaps with breast cancer dysregulated lncRNAs had a more significant enrichment rate, thereby identifying many new significantly enriched entries.
[0110] This method showed unique advantages in studying the transcription factors associated with lncRNA dysregulation in breast cancer. Compared with LncSEA2.0, this method was able to identify some new transcription factors that LncSEA2.0 failed to discover (P<0.01), as shown in Table 4. For the unique transcription factors in the enrichment analysis results of this method, their relevance to the occurrence and development of breast cancer was verified by literature retrieval and expression level change fold, which further demonstrated the accuracy of the enrichment analysis results of this method.
[0111] Table 4 Newly discovered transcription factors associated with breast cancer by TWLSEA
[0112]
[0113] In summary, the enrichment analysis results of the method of the present invention are more comprehensive and accurate. On the one hand, the heterogeneous networks related to lncRNA are fused to obtain a more comprehensive lncRNA-lncRNA functional similarity network; on the other hand, the importance scores of lncRNA are calculated, fully considering the different regulatory effects of different lncRNAs in different biological processes.
[0114] The above-mentioned documents are from:
[0115] [1] Chen
[0116] [2]Li J, Zhao Y, Zhou S, et al. Inferring lncRNAFunctional SimilarityBased on Integrating Heterogeneous Network Data[J].Frontiers inBioengineering and Biotechnology, 2020,8.
[0117] [3]Zhang G, Song C, Fan S, et al. LncSEA2.0: an updated platform for longnon-coding RNA related sets and enrichment analysis. Nucleic acids research2024.
[0118] [4]Zeng P, Chen J, Meng Y, et al. Defining Essentiality Score of Protein-Coding Genes and Long Noncoding RNAs. Frontiers in genetics 2018.
[0119] [5]Das T,Deb A,Parida S,et al.LncRBase V.2:an updated resource formultispecies lncRNAs and ClinicLSNP hosting genetic variants in lncRNAs forcancer patients.RNAbiology2021.
[0120] [6]Huang J,Zheng W,Zhang P,et al.ChIPBase v3.0:the encyclopedia oftranscriptional regulations of non-coding RNAs and protein-codinggenes.Nucleic acids research 2023.
[0121] [7]Bhartiya D,Pal K,Ghosh S,et al.lncRNome:a comprehensiveknowledgebase of human long noncoding RNAs.Database:the journal of biologicaldatabases and curation 2013.
[0122] [8]Zhang G,Song C,Fan S,et al.LncSEA2.0:an updated platform for longnon-coding RNA related sets and enrichment analysis.Nucleic acids research2024.
[0123] [9]Cheng L,Wang P,Tian R,et al.LncRNA2Target v2.0:a comprehensivedatabase for target genes of lncRNAs in human and mouse.Nucleic acidsresearch 2019.
[0124]
[10] Li J, Li Z, Wang Y, et al. TLSEA: a tool for lncRNA set enrichment analysis based on multi-source heterogeneous information fusion. Frontiers ingenetics 2023.
[0125] Any matters not described in the present invention are applicable to the prior art.
Claims
1. A lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance, characterized in that: The steps include: S1. Obtain lncRNA functional annotation set; S2. Construct lncRNA functional similarity matrices based on lncRNA-disease association, lncRNA-miRNA association and expression profile, respectively, and construct lncRNA feature vectors based on these matrices; calculate the Pearson correlation coefficient between any two lncRNA feature vectors, take lncRNA as a node, and the Pearson correlation coefficient as the edge weight to construct a lncRNA-lncRNA functional similarity network; The list of lncRNAs to be analyzed was expanded according to the lncRNA-lncRNA functional similarity network; S3. Select lncRNA importance attributes for the biological process of enrichment analysis, standardize the importance attribute values of each lncRNA, and obtain the importance score of each lncRNA; S4. Calculate the sum of the importance scores of all lncRNAs in the corresponding set according to formula (10), and calculate the distribution probability P of the lncRNA list on the lncRNA functional annotation set according to formula (11); Where N′ is the sum of the importance scores of all lncRNAs in the background lncRNA set N covered by the functional enrichment analysis, n′ is the sum of the importance scores of all lncRNAs in the lncRNA functional annotation set n, M′ is the sum of the importance scores of all lncRNAs in the lncRNA list M, k′ is the sum of the importance scores of all lncRNAs in the intersection of the lncRNA list and the lncRNA functional annotation set, is the importance score of the i-th lncRNA, and Q is the regulatory factor; If the distribution probability is less than or equal to the set threshold, it indicates that there is significant enrichment between the lncRNA list and the lncRNA functional annotation set, otherwise there is no significant enrichment.
2. The lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance according to claim 1, characterized in that: The process of constructing the lncRNA functional similarity matrix based on lncRNA-disease association is as follows: Define DG(u) and DG(v) as the associated disease sets associated with lncRNA u and v collected from the lncRNA-disease association data, and calculate the disease coefficient S(d) associated with lncRNAu in the associated disease set DG(v) according to the following formula: u , DG(v)) and the disease coefficient S(d v , DG(u)); Where, d u d v denote the diseases associated with lncRNA u and v, respectively, and DSS(·) represents the semantic similarity between diseases; The similarity S of the associated disease set DG(u) relative to DG(v) is calculated according to the following formula: u→v And the similarity S of the associated disease set DG(v) relative to DG(u) v→u : The functional similarity FS(u, v) of lncRNAu and v based on lncRNA-disease association was calculated according to the following formula: In the formula, |·| represents the total number of diseases in the associated disease set; According to the above process, the functional similarity of any two lncRNAs based on the lncRNA-disease association is calculated, and the functional similarity is used as a matrix element to construct a lncRNA functional similarity matrix based on the lncRNA-disease association.
3. The lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance according to claim 2, characterized in that: The semantic similarity calculation process between diseases is: The lncRNA associated with disease A and the corresponding terms in the medical disease vocabulary form a directed acyclic graph DAG A , DAG A The semantic contribution value of each term in to disease A is calculated by the following formula: In the formula, is the semantic contribution of term t to disease A, Δ is the weight of the edge connecting term t and t′, is the semantic contribution of term t′ to disease A, C(t) is the subset of term t, P t is the specific contribution factor of term t; The semantic value SV(A) of disease A is calculated by the following formula: Where, T A is the term set of disease A; Based on the intersection of the term sets of diseases A and B, the semantic similarity DSS(A, B) between diseases A and B is calculated; In the formula, is the semantic contribution of term t to disease B, SV(B) is the semantic value of disease B, T B is the term set of disease B.
4. The lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance according to claim 1, characterized in that: The process of constructing the lncRNA functional similarity matrix based on lncRNA-miRNA association is as follows: The miRNAs related to lncRNA u and v are listed in miRNA_list u and miRNA_list v , take the intersection of the two lists to get the list miRNA_list; construct a multidimensional vector based on the list miRNA_list to get the first miRNA associated feature vector Vector of lncRNA u and v u _1 and Vector v _1; miRNA_list u The set K of all miRNAs in u , list miRNA_list v The set K of all miRNAs in v , find the set K from the miRNA function similarity matrix u For K v The similarity between all miRNAs in lncRNA u is obtained, and the second miRNA associated feature vector Vector is obtained. u_2 Similarly, we obtain the second miRNA-associated feature vector Vector of lncRNA v v_2 ; The first miRNA associated feature vector of lncRNAu u_1 Vector associated with the second miRNA u_2 Connect the two ends to get the miRNA-associated feature vector Vector of lncRNA u u ; Similarly, we get the miRNA-associated feature vector Vector of lncRNA v v ; According to formula (8), the two miRNA association feature vectors Vector u and Vector v The cosine similarity Sim(lncRNAu, lncRNAv) is used as the functional similarity of lncRNAu and v based on the lncRNA-miRNA association; The above process was repeated to calculate the functional similarity of any two lncRNAs based on lncRNA-miRNA association, and the lncRNA functional similarity matrix based on lncRNA-miRNA association was constructed.
5. The lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance according to claim 1 or 4, characterized in that: The process of constructing the lncRNA functional similarity matrix based on expression profile is as follows: construct an expression profile feature vector for each lncRNA, calculate the Pearson correlation coefficient between the expression profile feature vectors of two lncRNAs, use this coefficient as the functional similarity of the two lncRNAs based on expression profile, and construct the lncRNA functional similarity matrix based on expression profile.
6. The lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance according to claim 1, characterized in that: The process of constructing the lncRNA feature vector is as follows: for lncRNA that only appears in one of the functional similarity matrices, the row vector of the lncRNA is extracted from the corresponding functional similarity matrix as the feature vector; for lncRNA that appears in two or three functional similarity matrices at the same time, the row vector of the lncRNA is extracted from the corresponding functional similarity matrix respectively, and the three row vectors are concatenated to obtain the feature vector of the lncRNA.
7. The lncRNA functional enrichment analysis method based on heterogeneous network fusion and importance according to claim 1, characterized in that: The amplification process of the lncRNA list to be analyzed is as follows: setting the similarity coefficient, taking all lncRNAs in the lncRNA list to be analyzed as seed nodes, discarding the edges with weights less than the similarity coefficient in the lncRNA-lncRNA functional similarity network, and amplifying by random walk.