Method for predicting activity of ubiquitin ligase and deubiquitination enzyme

Abstract

The invention discloses a method for predicting activity of ubiquitin ligase and deubiquitination enzyme. The method comprises the following steps: 1) collecting a real interaction set from an experimental verification database; merging the data set with a prediction interaction data set to obtain an integrated prediction set; 2) respectively matching quantitative ubiquitination omics data provided by a user with the real interaction set and the integrated prediction set, and extracting a Log2 difference multiple Log2 FC value in the quantitative ubiquitination omics data from matched entries; 3) calculating the enzyme activity of the real interaction set; 4) performing dynamic threshold optimization and screening on the integrated prediction set to obtain a comprehensive score; (5) selecting a threshold value which maximizes the comprehensive score as an optimal prediction confidence threshold value, and then screening out interaction data of which the confidence score is higher than the optimal prediction confidence threshold value from the integrated prediction set, (6) calculating the enzyme activity of the prediction set, and (7) constructing an enzyme activity prediction report based on results obtained in the steps (3) and (6).

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of computational systems biology and functional proteomics, specifically relating to a method for predicting the activities of ubiquitin ligases and deubiquitylases, used to systematically analyze the regulatory dynamics of the ubiquitin-proteasome system from high-throughput proteomics data. This invention particularly focuses on developing a proprietary weighted substrate enrichment analysis framework that, by integrating validated and predicted protein-protein interaction networks, quantitatively infers and statistically evaluates the functional activities of ubiquitin ligases (E3 ubiquitin ligase) and deubiquitylases (DUB) under specific pathophysiological perturbations (such as disease states, drug interventions, genetic modifications, etc.). Background Technology

[0002] Ubiquitylation plays a central role in regulating core cellular life activities such as protein stability, subcellular localization, activity, and interactions. The dynamic process of ubiquitination is catalyzed by a cascade reaction consisting of ubiquitin activators (E1), ubiquitin conjugates (E2), and ubiquitin ligases (E3), and is inversely regulated by deubiquitinases (DUB). E3 and DUB act as the "writing" and "erasing" elements of this system, respectively, maintaining the precise balance of the intracellular ubiquitination network through their substrate specificity. With advancements in mass spectrometry, researchers have been able to identify ubiquitination modification sites and their abundance that change under different physiological or pathological conditions on a large scale using affinity enrichment strategies (e.g., using ubiquitin-binding domain tools TUBEs, ubiquitin chain-specific antibodies, or chemical ubiquitin probes) combined with high-throughput mass spectrometry analysis. However, how to inversely and accurately infer changes in the functional activity of upstream E3 / DUB regulatory elements from these downstream, complex ubiquiome data remains a core computational biology challenge in this field. Existing analytical methods face several key technical bottlenecks in this transformation process: 1. Statistical models for inferring enzyme functional activity lack directionality. Current computational tools largely focus on E3 enrichment analysis, lacking a systematic assessment of DUB function. Crucially, existing frameworks fail to fully integrate the fundamental opposition between E3 and DUB in their regulatory logic: enhanced E3 activity is expected to lead to increased substrate ubiquitination levels (positive correlation), while enhanced DUB activity leads to decreased substrate ubiquitination levels (negative correlation). The lack of such a bidirectional and oppositely oriented integrated statistical model makes it impossible to accurately analyze the complex regulatory dynamics within the ubiquitination network.

[0003] 2. The dual dilemma of the "quality" and "quantity" of interaction data. Reliable enzyme-substrate interactions are a prerequisite for activity inference. However, the number of experimentally validated, high-quality E3 / DUB-substrate pairs (real interactions) is currently limited, severely restricting the scope and sensitivity of the analysis. Although various computational methods (based on domain, linear ordering, co-evolution, or network topology features) predict a large number of potential interactions, these predicted datasets generally suffer from heterogeneity in reliability. Indiscriminate use in the analysis will introduce significant noise, affecting the accuracy of the inference results.

[0004] 3. Traditional enrichment analysis methods have limitations when integrating heterogeneous data. Classical substrate set enrichment analysis methods (such as unweighted Z-score tests or gene set enrichment analysis GSEA) perform reasonably well when processing validated interaction datasets. However, when attempting to integrate predicted interaction data with a wide confidence range, these methods fail to assign higher weights to high-confidence interactions, leading to the analysis results being diluted or misleading by a large number of low-confidence predictions, making it difficult to accurately reflect the enzyme's activity state. Summary of the Invention

[0005] To address the technical problems existing in the prior art, the purpose of this invention is to provide a method for predicting the activities of ubiquitin ligases and deubiquitinating enzymes, which can simultaneously cover the bidirectional regulatory logic of E3 and DUB, and can intelligently integrate and weight multi-source heterogeneous interaction data to achieve a more reliable and comprehensive inference from ubiquitination data to changes in upstream regulatory enzyme activity.

[0006] The technical solution of this invention is as follows: A method for predicting ubiquitin ligase (E3) and deubiquitinase (DUB) activities, comprising the following core steps.

[0007] 1) Integration and Preprocessing of Multi-Source Interaction Data: Real E3-substrate and DUB-substrate interaction datasets (E3real, DUBreal) were collected from the experimental validation database, along with predicted E3-substrate and DUB-substrate interaction datasets from the prediction model, accompanied by confidence scores (interScore). Subsequently, the real and predicted sets were merged to construct an integrated prediction set (E3pre, DUBpre). The merging rules are as follows: Data Merging: The real interaction set (E3real, DUBreal) and the predicted interaction datasets were fully connected and merged to form the initial integrated prediction set (E3pre, DUBpre); Overlapping Entries Handling: During the merging process, if an entry in the prediction dataset overlaps with an entry in the real interaction set (E3real, DUBreal) (i.e., the enzyme-substrate pair is identical), then that overlapping entry is considered a validated real interaction in the integrated prediction set. Confidence score assignment: Assign a uniform confidence score to each entry in the integrated prediction set (E3pre, DUBpre) according to the following rules: For entries originating from the true interaction set, or entries that overlap with the true set, set their confidence score (interScore) to the maximum value of 1; for entries that exist only in the prediction dataset and do not overlap with the true set, retain their original prediction confidence score (interScore).

[0008] 2) Experimental data matching and fold change (FC) extraction: The quantitative ubiquitination data provided by the user (including the UniProtKB Accession of the substrate protein and the Log2 fold change Log2 FC value) are matched with the above real interaction set (E3real, DUBreal) and integrated prediction set (E3pre, DUBpre) using the substrate UniProtKB Accession. The Log2 FC value corresponding to the specific ubiquitination site recorded in the known interactions (if applicable) is then extracted.

[0009] 3) Enzyme activity calculation from real interaction sets (standard Z-score method): For real E3-substrate and DUB-substrate interaction datasets (E3real, DUBreal), calculate the average Log2 FC value of the substrate site corresponding to each enzyme (E3 or DUB). ) and number of substrate sites ( ), calculate the overall mean of the Log2 FC values ​​for all substrate sites ( ) and population standard deviation ( The activity scores of E3 and DUB enzymes were calculated using standard Z-score enrichment analysis, with the Z-score calculation direction for E3 being opposite to that for DUB. Then, a two-tailed test was used to calculate the original value of each enzyme activity change. Value. The calculation formula is: To control for false positives in multiple hypothesis testing, the Benjamini-Hochberg (BH) method was used independently to correct the false discovery rate (FDR) for the p-value sets calculated for all E3 enzymes and the p-value sets calculated for all DUB enzymes, thereby calculating the corresponding q-value (i.e., the FDR correction value) for each p-value in the set. Finally, for each E3 enzyme, its specificity Z-score is output. ), p-value ( ) and FDR correction value ( ); output the specificity Z-score for each DUB enzyme ( ), p-value ( ) and FDR correction value ( ).

[0010] 4) Dynamic threshold optimization and screening of the integrated prediction set: For the integrated prediction set (E3pre, DUBpre), a series of thresholds are set within the confidence score range (0.5 to 1.0). At each threshold, interaction entries with an interScore higher than that threshold are screened, and the number of enzyme types at the current threshold is calculated. ) and total number of interactions ( ), by and After normalization and weighted combination, the overall score is obtained: This is used to evaluate the analysis coverage and data volume at this threshold.

[0011] 5) Select the threshold that maximizes the CompositeScore as the optimal prediction confidence threshold, and based on this optimal prediction confidence threshold, filter the interaction data from the integrated prediction set (E3pre, DUBpre) whose interScore is higher than this optimal prediction confidence threshold.

[0012] 6) Enzyme activity calculation of the integrated prediction set (weighted Z-score method): For the E3pre and DUBpre sets obtained in step 5), the confidence scores of each interaction in the filtered data are first normalized to obtain the normalized weights: This step ensures that all weights fall within the [0, 1] interval. For each enzyme (E3 or DUB), its specific weighted statistic is calculated. Weighted average Log2 FC: Total weights: Weighted population standard deviation: Enzyme activity was calculated using weighted Z-score enrichment analysis, with the DUB direction opposite to that of E3. E3 ligase: DUB deubiquitinase: Statistical significance assessment: Statistical significance was calculated based on weighted Z-scores. E3 ligase: DUB deubiquitinase: To control for false positives in multiple hypothesis testing, independent FDR corrections were performed on the p-value sets of E3 and DUB, respectively. E3 ligase: DUB deubiquitinase: .

[0013] The following is a description of the parameters: : Normalized weights of the i-th enzyme and the j-th substrate site; The weighted average Log2 FC value of the enzyme at the corresponding substrate site; : The sum of normalized weights of all substrate sites of the enzyme; : The weighted population mean of all matched substrate sites; : Weighted population standard deviation of all matched substrate sites; Weighted activity score of E3 / DUB enzyme; : The original p-value of the change in E3 / DUB enzyme activity; FDR correction value for changes in E3 / DUB enzyme activity.

[0014] Calculation instructions: This represents the summation over all substrate sites corresponding to a single enzyme; This represents the summation over all matching substrate sites (across all enzymes); The cumulative distribution function represents the standard normal distribution; This indicates that the Benjamini-Hochberg method is used for error detection rate correction.

[0015] 7) Construct a complete enzyme activity prediction report based on the enzyme activity calculations obtained from step 3) (real interaction set) and step 6) (integrated prediction set). Activity strength is determined by the Z-score value for each enzyme (including...). and This represents the quantitative prediction result of enzyme activity. A positive value indicates a significant increase in enzyme activity relative to the baseline under experimental conditions, while a negative value indicates a significant decrease in enzyme activity relative to the baseline under experimental conditions. The absolute value reflects the magnitude of the activity change. To control for false positives, the FDR-corrected q-value is recommended as the screening criterion. For the true interaction set results, based on... and Perform filtering; for the integrated prediction set results, based on Screening is performed, typically using q<0.05 or q<0.1 as the threshold for statistical significance. The final output is a comprehensive activity report that includes enzyme identifier (UniProtKB Accession, gene symbol), Z-score, p-value, q-value, and data source (real set / integrated prediction set).

[0016] Application Guidance: Results Integration and Biological Interpretation: Cross-validate enzyme activity results from the real set and the integrated prediction set. Prioritize enzymes that show significant activity changes (e.g., q < 0.05) in both datasets and with consistent direction of change. Enzymes that are significant only in a single dataset require careful evaluation considering their biological function and experimental context. This activity prediction result can be used to identify E3 ligases and DUBs that are critically regulated under specific pathophysiological conditions (e.g., disease states, drug treatments, etc.). Enzymes with significant activity changes can serve as potential drug targets or biomarkers, providing priority directions for subsequent functional validation experiments. By comparing activity change patterns under different experimental conditions, the dynamic characteristics of the ubiquitination regulatory network can be further analyzed. The final enzyme activity prediction report provides reliable computational evidence and priority screening criteria for downstream biological mechanism research and experimental validation.

[0017] The present invention has the following advantages.

[0018] 1. Improved prediction accuracy and robustness: A weighted Z-score statistical model based on prediction confidence (interScore) is introduced, which can accurately reflect the influence weight of high-confidence interacting substrates on enzyme activity, rather than a simple average, thereby significantly improving the accuracy and statistical robustness of enzyme activity prediction based on integrated data.

[0019] 2. Intelligent optimization of prediction data: An innovative dynamic threshold evaluation mechanism is introduced. By optimizing the combined score of the number of unique enzymes and the total number of interactions, the optimal prediction interaction screening threshold is automatically determined, solving the subjectivity and blindness problems of manually setting thresholds by users. The curve of the dynamic threshold evaluation for E3 ligase prediction interactions is shown below. Figure 1 As shown, Figure 1 The two curves above show the trends of the unique E3 enzyme number score and the total number of interactions score as a function of the confidence threshold, respectively. Figure 1The vertical dashed line in the graph represents the optimal confidence threshold recommended after comprehensive score optimization. The curve for evaluating the dynamic threshold of DUB deubiquitinase predictive interactions is shown below. Figure 2 As shown, Figure 2 The two curves above show the trends of the unique DUB enzyme number score and the total number of interactions score as a function of the confidence threshold, respectively. Figure 2 The vertical dashed line in the figure represents the optimal confidence threshold recommended after optimization by comprehensive score.

[0020] 3. A unified two-way analysis framework is provided: In the same analysis process, the Z-score calculation methods with opposite directions are used for the action mechanisms of E3 ligase (activation) and DUB deubiquitinase (inhibition), realizing integrated and efficient analysis of E3 and DUB activities.

[0021] 4. Strong data compatibility and traceability: It supports input of ubiquitomics data in multiple formats and can simultaneously output results based on real data (E3 / DUB_real_activity.csv) and integrated prediction data (E3 / DUB_pre_activity.csv), which facilitates researchers to compare and trace the results. Attached Figure Description

[0022] Figure 1 This is a graph showing the dynamic threshold assessment of predicted interactions using the E3 ligase in this invention.

[0023] Figure 2 This is a graph showing the dynamic threshold assessment of the predicted interaction between DUB deubiquitinating enzymes in this invention.

[0024] Figure 3 This is a flowchart of the enzyme activity prediction method of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand and implement the present invention, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The present invention proposes a method and system for predicting ubiquitination / deubiquitination enzyme activity based on weighted substrate enrichment analysis. This method and system can be deployed on computing devices with corresponding computing capabilities (such as servers and personal computers). It is mainly implemented through the R language environment and relies on specific bioinformatics databases (such as org.Mm.eg.db, org.Hs.eg.db, etc.) and data processing libraries (such as dplyr, data.table, clusterProfiler, etc.).

[0026] The data used in this study were derived from a self-produced database of ubiquitination and proteomics data from human hepatocellular carcinoma. The aim was to investigate the differences in enzyme activity at the level of protein ubiquitination modification between tumor tissues and their corresponding adjacent normal tissues during the development and progression of hepatocellular carcinoma (HCC).

[0027] First, data filtering was performed to retain protein / modification rows with a non-missing value ratio ≥ 2 / 3. Then, quantile normalization was applied to the filtered data to eliminate systematic technical biases between samples (such as differences in sample loading amount and instrument sensitivity). For missing values ​​still present after normalization, global minimum values ​​were used for imputation. The imputated data underwent a base-2 logarithmic transformation. Finally, Ubi-Pro was used based on the protein ID shared by the ubiquitination group (Ubi) and proteome (Pro) of each group. median The new-Ubi was obtained; differential expression analysis of the new-Ubi was performed using the limma package (Linear Models for Microarray Data).

[0028] Comparison settings: Comparison 1: Early_Normal vs Early_Tumor (identifying early-stage HCC tumors - normal differences); Comparison 2: Advanced_Normal vs Advanced_Tumor (identifying late-stage HCC tumors - normal differences); Comparison 3: Early_Normal + Advanced_Normal vs Advanced_Normal + Advanced_Tumor (identifying overall HCC tumors - normal differences); Enzyme activity analysis was performed based on the above three comparisons.

[0029] Human species interaction data matching (corresponding) Figure 3 This paper outlines the complete process of data integration, real / predicted set branching, standard Z-score calculation, dynamic threshold optimization, weighted Z-score calculation, and final result output, and performs substrate UniProtKB Accession and ubiquitination modification site matching on the four interaction sets E3_real and DUB_real, E3_pre and DUB_pre.

[0030] Enzyme activity calculation for real interaction sets (standard Z-score method): Enzyme activity was calculated based on the results of matching the E3_real and DUB_real sets with substrate omics data. Local mean calculation: For each enzyme... Calculate the overall statistic: Calculate the statistic for all substrate sites. Calculate the Z-score: Calculate the enzyme activity Z-score according to the standard Z-score enrichment analysis formula. Statistical significance: Calculate the p-value (using a two-tailed test) and perform FDR correction.

[0031] The dynamic threshold optimization and weighted Z-score calculation of the integrated prediction set are performed. For E3_pre and DUB_pre, dynamic threshold optimization is performed using interScore as the weight: the confidence score iteration range is 0.5 to 1.0. At each threshold, the number of unique enzymes is calculated. and total number of interactions The composite score (CompositeScore) was used to select a threshold of 0.72 that maximizes the CompositeScore as the optimal prediction confidence threshold, and interaction data were filtered based on this optimal threshold. Weighted Z-scores were calculated, p-values ​​were calculated, and FDR correction was performed.

[0032] The system integrates all calculated enzyme activity results (Z-score, p-value, FDR) and performs a mapping conversion from UniProtKB Accession to gene symbols to output the final enzyme activity results. The results show that the activities of E3 enzyme Socs2 and DUB enzyme Usp21 decreased in HCC tumor tissue, while the activities of E3 enzymes E6 and Fbxw7 and DUB enzyme Atxn3 increased in HCC tumor tissue. These ubiquitinizing enzymes may be potential drug targets, pointing the way for further experimental validation.

[0033] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the activities of ubiquitin ligases and deubiquitinating enzymes, comprising the steps of: 1) Collect real E3-substrate and DUB-substrate interaction datasets from the experimental validation database, denoted as the real interaction set (E3real, DUBreal); merge the real interaction set (E3real, DUBreal) and the predicted interaction dataset to obtain the integrated prediction set (E3pre, DUBpre). 2) Based on the universal protein knowledge base login number of the substrate, the quantitative ubiquitination omics data provided by the user is matched with the real interaction set (E3real, DUBreal) and the integrated prediction set (E3pre, DUBpre), respectively. From the successfully matched entries, the Log2 difference fold Log2 FC value of the quantitative ubiquitination omics data is extracted. 3) Calculate the average Log2 FC value for each enzyme at the substrate site in the real interaction set (E3real, DUBreal). and number of substrate sites Calculate the overall mean of the Log2 FC values ​​for all substrate sites. and overall standard deviation Then, the activity scores of ubiquitin ligase E3 were calculated using the standard Z-score enrichment analysis method. Activity score of deubiquitinase DUB Then, a two-tailed test was used to calculate the original value of each ubiquitin ligase E3. value The original DUB of each deubiquitinase value And for E3, DUB The error detection rate (FDR) is corrected by applying an error detection rate correction (FDR) to the set of values, resulting in the corresponding FDR correction value. , ; 4) For the integrated prediction set (E3pre, DUBpre), set a series of thresholds within the confidence score range; at each threshold, select interaction entries with interaction scores higher than that threshold, and calculate the number of enzyme types at that threshold. and total number of interactions By and Normalization and weighted combination are performed to obtain the composite score, which is used to evaluate the analysis coverage and data volume at this threshold. 5) Select the threshold that maximizes the CompositeScore as the optimal prediction confidence threshold, and then filter the interaction data with confidence scores higher than the optimal prediction confidence threshold from the integrated prediction set (E3pre, DUBpre). 6) Normalize the confidence scores of each interaction data selected in step 5) to obtain the normalized weights of the i-th enzyme and the j-th substrate site. Then, based on the weighted statistics for each enzyme, including the weighted average Log2 FC, the total weight, and... Weighted population standard deviation Enzyme activity was calculated using weighted Z-score enrichment analysis to obtain the enzyme activity of E3. DUB enzyme activity ;Statistical significance assessment of E3 Then, the error detection rate (FDR) is corrected, and the corresponding FDR correction value is obtained. ;Significance assessment of statistical DUB And perform error detection rate (FDR) correction on it, with the corresponding FDR correction value. ; 7) Construct an enzyme activity prediction report based on the results obtained in steps 3) and 6).

2. The method according to claim 1, characterized in that, In step 3), ; The primitive ubiquitin ligase E3 value The original deubiquitinase DUB value ; The cumulative distribution function represents the standard normal distribution; ubiquitin ligase E3 value Deubiquitinase DUB value .

3. The method according to claim 1, characterized in that, In step 4), the overall score is calculated. .

4. The method according to claim 1, characterized in that, In step 6), ; ; , , ; .

5. The method according to claim 1, characterized in that, In step 7), the quantitative prediction results of enzyme activity are directly characterized by the Z-score value of each enzyme; the enzyme activity prediction report includes enzyme identifier, Z-score, p-value, q-value and data source; the enzyme identifier includes the UniProtKB Accession of the substrate protein and the gene symbol; the quantitative ubiquitination data provided by the user are matched with the real interaction set (E3real, DUBreal) and the integrated prediction set (E3pre, DUBpre) according to the substrate UniProtKB Accession, and the Log2 fold change value (Log2 FC value) is extracted from the successfully matched entries in the quantitative ubiquitination data.

Images