Nucleolar Knowledge Base (NKB) User Guide

Table of Contents

1. Knowledge Base Overview
2. Quick Start
3. Query Methods Guide
4. Detailed Operation Guide
5. Advanced Features
6. Frequently Asked Questions
7. Technical Specifications
8. Contact Us

1. Knowledge Base Overview

1.1 What is NKB?

The Nucleolus Knowledge Base (NKB) is a comprehensive knowledge and potential insight repository focused on nucleolar biology and related diseases, built from literature databases including PubMed.

1.2 What information does NKB provide?

• Evidence-based knowledge and potential insights in the precise form of Triples, Subject-Predicate-Object (SPO), labelled with HIGH/MIDDLE/LOW CONFIDENCE
• Evidence-based mini-review based on all obtained Triples or the HIGH CONFIDENCE ones selected from confidence screening
• Knowledge graphs presented in a visualized tree structure display the hierarchical relationships and associations between knowledge entities

1.3 Core Data

1.4 Main Features

• Structured Knowledge Retrieval: Evidence-based conclusions in SPO triple format
• AI Review Generation: Automatically generate mini-reviews with citations based on literature
• Knowledge Graph Visualization: Hierarchical directory structure for knowledge discovery
• Three-Level Confidence Evaluation: HIGH/MIDDLE/LOW

2. Quick Start

➢ 3-Minute Quick Experience (for First-Time Users)

(1) Enter Homepage → Click Browse Triples → The following new page with 22252 Triples listed for the entire Knowledge Base.

Each triple is equipped with a label of HIGH/MIDDLE/LOW CONFIDENCE on the right top, so that you can find the high confidence Triples that are based on solid facts.

On the right bottom, you can click the Details for more detailed information for this current Triple.

(2) Enter keywords (e.g. FBL) in the query box → Click Search

A subset of keyword-related Triples listed at the bottom.

New searching box appears with button Synthesize on the right.

(3) Enter a question in the query box, e.g. "what is the function of FBL in cancer cells" → click Quick Filter → Select the desired confidence range (If want to view all confidence level entries, click Clear Filter) → click Synthesize → "Synthesize" turned into "Thinking" → Wait for minutes → Receive a 200-500 word literature review with PMID citations

3. Query Methods Guide

3.1 Three query methods in the home page

3.2 How to Choose the Right Query Method?

3.3 Navigation Bar Functions

4. Detailed Operation Guide

4.1 Browse Triples

Steps:

(1) Enter Homepage → Click "Browse Triple" → The following new page with 22252 Triples listed for the entire Knowledge Base.

Each triple is equipped with a label of HIGH/MIDDLE/LOW CONFIDENCE on the right top, so that you can find the high confidence Triples as potential valuable insights.

On the right bottom, you can click the "Details" for more detailed information for this current Triple.

(2) Enter keywords in the search box, e.g FBL → Click "Search" → A subset of FBL-related Triples listed at the bottom.

New searching box appears with "Synthesize" button on the right.

Use Quick Filter on the Navigation bar to remove Triples with a certain Confidence level

(3) Enter a keyword-related question in the query box, e.g. "what is the function of FBL in cancer cells".

→ click "Quick Filter" → Select the desired confidence range (If want to view all confidence level entries, click "Clear Filter") → click "Synthesize" → "Synthesize" turns into "Thinking". Please wait for minutes → Receive a 200-500 word literature review with PMID citations

4.2 Knowledge Graph

Steps:

(1) Enter Homepage → Click "Knowledge Graph"

(2) Select Category → Choose relevant main category in tree structure

(3) Browse Subcategories → Click to expand subcategories and related entries

(4) Get Entry → Select entry to retrieve all related SPO triples

(5) Ask Question → Enter specific question

(6) Click "Quick Filter" → Select the desired confidence range (If want to view all confidence level entries, click "Clear Filter")

(7) Generate Review: Click "Synthesize" to get review

4.3 Generate Search Query

Steps:

(1) Enter Homepage → Row down the page and see the query box → Input Question → Click "Generate Search Query"

(2) The system automatically transfers the input question into keywords

(3) Select Keywords: Choose relevant keywords from recommendations → Click "Immediate Search"

(4) Get Results: System retrieves corresponding knowledge entries

(5) Optimize Keywords: If no results → filter or refine keywords and copy

(6) Repeat Search: Paste the copied keywords into the search box → click "Search"

(7) Generate Review: After successful retrieval, enter question

(8) Click "Quick Filter" → Select the desired confidence range (If want to view all confidence level entries, click "Clear Filter")

(9) Click "Synthesize" → Get review

5. Advanced Features

5.1 AI Review Generation System

Function: Automatically generate 200-500 word academic reviews based on retrieved knowledge entries

Features:

• ✅ Automatic PMID citations included
• ✅ One-click copy functionality
• ✅ Based on evidence-based knowledge
• ✅ Academic writing standards

Usage Recommendations:

• Questions should be specific and clear
• Avoid overly broad inquiries
• Can iterate multiple times to optimize question description

5.2 Confidence Scoring System

5.3 Quick Filter

Function: Preset confidence filters supporting multi-criteria rapid filtering

Use Cases:

• Large volume of results need quick filtering
• Focus only on high-quality evidence
• Filter results by specific criteria

6. Frequently Asked Questions

Q1: What to do when search returns no results?

A:

1. Check keyword spelling
2. Try synonyms or simplified terms
3. On the homepage, use "Generate Search Query" for recommended keywords
4. Try broader search terms

Q2: How to improve search accuracy?

A:

1. Use standard gene/protein names
2. Combine with confidence filtering
3. Check "Details" for complete information
4. Cross-validate multiple sources

Q3: How is AI review quality ensured?

A:

1. Based on high-quality literature databases
2. Includes original PMIDs for verification
3. Confidence scoring assists judgment
4. Recommend combining with professional knowledge

Q4: What is the data update frequency?

A:

• Regular updates from PubMed and other databases
• Check "About" page for specific update cycles

Q5: Can data be exported?

A:

• Reviews support one-click copying
• Contact administrator for detailed data export

Q6: What to do with technical issues?

A:

1. Refresh page and retry
2. Check network connection
3. Review other FAQ answers
4. Contact technical support: niyx@szu.edu.cn

7. Technical Specifications

7.1 Data Sources

• Primary Source: PubMed literature database
• Data Types: SPO triples, gene information, protein functions, disease associations, pathway, etc.
• Update Mechanism: Regular automatic updates

7.2 System Architecture

• Knowledge Extraction: Based on natural language processing technology
• Confidence Calculation: Multi-dimensional evidence strength assessment
• Search Engine: Supports keyword and semantic search
• AI Generation: Based on retrieval-augmented text generation

7.3 Usage Limitations

• Academic research use only
• Not for commercial purposes
• Proper source citation required

8. Contact Us

Technical Support

📧 Email: niyx@szu.edu.cn

🏢 Institution: Shenzhen University

Feedback and Suggestions

We welcome your feedback and suggestions! Please contact us via:

• Send user experience and improvement suggestions
• Report system bugs or technical issues
• Request new features or data updates
• Collaboration and exchange opportunities

Citation Information

If NKB helps your research, please cite appropriately in publications. Check the "About" page for specific citation information and format.

Appendix: Glossary