pi-USEA: Ubiquitin Ligase/Deubiquitase Substrate Enrichment Analysis

1. Tool Overview

This tool is specifically designed to predict the activity of ubiquitin ligases and deubiquitinases. Users only need to upload the ratio change data of their ubiquitinome of interest to obtain corresponding enzyme activity prediction results.

2. Core Features

1. Dual Activity Prediction: Supports prediction of both ubiquitin (E3 ligase) and deubiquitinase (DUB) activity.
2. Flexible Input: Accepts data with or without site-specific modification information.
3. Dual Prediction Modes: Supports analysis based on validated interaction data, and can also integrate predicted interaction data for weighted combined calculations.
4. Multi-Species Support: Pre-loaded reference datasets for human (Homo sapiens) and mouse (Mus musculus).
5. Validated Performance: Rigorously tested on benchmark datasets, ensuring reliable results.

3. Usage Workflow

3.1 Environment Setup

• R Environment: Requires R version 3.5 or higher.
• Code and Dataset Download:
  • Code files: [Address]
  • Datasets: part1_interaction[Address], part2_calculation: Baidu Netdisk (Extraction Code: mp94)

3.2 Input Data Preparation

• File Format: Can be saved as test_data.csv, containing the following columns (at minimum, SUB_ACC_ID and FC are required):

Column Name	Description	Example
SUB_ACC_ID	UniProt ID of the substrate protein	A2A432
SYMBOL	Substrate gene symbol (optional)	Cul4b
SUB_MOD_RSD	Substrate modification site (optional)	K402p
P.Value	Adjusted p-value (optional)	0.411203714
FC	log2 ratio of experimental group/control group (required)	-4.376820087

• Example file: test_data.csv Example

3.3 Analysis Steps

The tool's datasets have been pre-processed for human and mouse species. If your data is from human or mouse, you can directly use the corresponding species dataset provided by the tool. Open pi-USEA.R, install and load the required R packages, and start the analysis directly from Part 2. The main steps are:

1. Read Interaction Datasets: Load the interaction datasets (including E3pre: combined predicted and real E3 interactions; E3real: real E3 interactions; DUBpre: combined predicted and real DUB interactions; DUBreal: real DUB interactions) and your prepared ubiquitinome data.
2. Merge Datasets: Merge the interaction datasets with your ubiquitinome data.
3. Analyze Real Datasets (E3real, DUBreal): Directly calculate enzyme activity and output results E3/DUB_real_activity.csv.
4. Analyze Predicted Datasets (E3pre, DUBpre): Perform confidence assessment, select a confidence threshold, and filter high-confidence interaction entries (manual threshold setting required) for subsequent analysis, for example: E3pre <- E3pre %>% filter(interScore > 0.75)
5. Weighted Calculation for Predicted Datasets: Calculate enzyme activity using a weighted average and output results.
   • E3_real_activity.csv / DUB_real_activity.csv: Activity results based on validated interactions.
   • E3_pre_activity.csv / DUB_pre_activity.csv: Activity results integrating predicted data.

4. Result Field Descriptions

4.1 Core Identifiers

Field Name	Description	Example
ENZY_ACC_ID	UniProt ID of the enzyme (E3 ubiquitin ligase or DUB)	P04637 (human p53 protein)
SUB_ACC_ID	UniProt ID of the substrate protein	Q13547 (histone deacetylase 1)
SUB_MOD_RSD.x	Modified residue position on the substrate protein (merged from SUB_MOD_RSD in test_data)	K48, K341, K363
ENZYME	Name of the enzyme	UBE3A (E3 ubiquitin ligase)
SUBSTRATE	Name of the substrate protein	TP53 (tumor suppressor protein p53)
GENE	Gene name corresponding to the enzyme	Ube3a
SUB_GENE	Gene name corresponding to the substrate	Hdac1
SYMBOL	Gene symbol (from test_data)	Hdac1

4.2 Experimental Data

Field Name	Description	Calculation
FC	Fold change (from test_data)	log2 (experimental group / control group)
logFC_mean	Average log2FC for all substrate sites corresponding to a single enzyme	mean(log2(FC))
logFC_count	Number of substrate sites corresponding to a single enzyme	Direct count

4.3 Statistical Significance

Field Name	Description	Formula
z_score	Quantitative result of enzyme activity. Positive indicates increased activity, negative indicates decreased activity.	Real: (logFC_mean - overall mean) / (overall sd / sqrt(n)) Predicted: (weighted logFC_mean - overall weighted mean) / (overall weighted sd / sqrt(weight)) Note: DUB direction is opposite.
p_value	p-value calculated from z_score	2 * pnorm(-abs(z_score)) (two-tailed test)
fdr	False discovery rate corrected p-value	Benjamini-Hochberg correction

4.4 Prediction Score Fields (This analysis primarily uses interScore. For details, refer to the UbiBrowser original publication.)

Field Name	Description	Score Range	Source
interScore	Enzyme-substrate interaction prediction score	0.5 - 1	UbiBrowser and other prediction tools
domainScore	Domain interaction score	-	InterPro and other domain databases
motifScore	Substrate motif matching score	-	e.g., linear ubiquitination motif matching
netScore	Network topology importance score	-	Centrality analysis based on PPI networks
goScore	Gene ontology functional similarity score	-	GO semantic similarity calculation

4.5 Source Markers

Field Name	Description	Typical Values
source	Data source (after merging)	UbiBrowser_real, E3Net, UbiBrowser_pre
source_pre	Predicted data source	UbiBrowser_pre
source.x / source.y	Redundancy columns for real vs. predicted interactions	x: UbiBrowser_pre y: UbiNet, UbiBrowser_real

4.6 Threshold Evaluation

Field Name	Description	Purpose
Threshold	interScore score threshold	Used to filter reliable interactions
num_unique_enzy	Number of unique enzymes at the current threshold	Assess coverage
num_rows	Total number of interactions at the current threshold	Assess data volume

4.7 Other Technical Fields

Field Name	Description
KIN_ORGANISM	Species of the enzyme
SUB_ORGANISM	Species of the substrate
exists_real	Indicates if the interaction is real (TRUE/FALSE)
normalized_interScore	Normalized interaction score ((interScore - min) / (max - min))

5. Custom Analysis (Non-Human/Mouse Species)

1. If you need to analyze species other than human and mouse, please download interaction data from suggested databases:

   • UbiBrowser
   • UbiNet
   • E3Net

2. Modify column names according to the format used in this tool's datasets:

   • Real interaction datasets: After downloading, modify column names to include ENZY_ACC_ID: UniProt ID of the enzyme (required), SUB_ACC_ID: UniProt ID of the substrate (required).
   • Predicted interaction datasets from UbiBrowser: Can be read directly from the downloaded .gz file.
   • After preparing the required file formats and names, install necessary packages and start the analysis from Part 1.

Contact Information

If you have any questions during usage, please feel free to contact us:

• Chang Cheng : changchengbio@163.com or changcheng@ncpsb.org.cn
• Liu Ning: liu_ning2021@163.com