📊 GraphAvalanche

Interactive Citation Network Visualizer for AVALANCHE Literature Discovery Results

Transform your AVALANCHE literature search into beautiful, interactive citation networks with automatic tier-based highlighting. Create Connected Papers-style visualizations for your systematic reviews and meta-analyses.

Python 3.8+ License: MIT Streamlit

• Connected Papers-style visualization - Navigate citation relationships intuitively
• Tier-based color coding - Automatically highlights high-impact papers
• Multiple layout algorithms - Spring, Kamada-Kawai, Circular, Spectral
• Interactive exploration - Zoom, pan, hover for details

• Automatic tiering - Papers categorized by citation impact (customizable thresholds)
• Network statistics - Density, connectivity, influential papers
• Timeline view - Chronological publication trends
• OpenAlex integration - Fetches citation relationships automatically

• GraphML export - Import to Gephi, Cytoscape, or igraph
• CSV export - Data analysis in R/Python/Excel
• Publication-ready figures - High-quality visualizations for papers

# Clone the repository
git clone https://github.com/HaroldMate1/GraphAvalanche.git
cd GraphAvalanche
# Install dependencies
pip install -r requirements.txt

python -m streamlit --version python -m streamlit run app_v2.py

First, run AVALANCHE to discover literature:

python avalanche.py 10.1038/s41586-020-2649-2

This creates avalanche_results.xlsx with your papers.

1. Launch GraphAvalanche: streamlit run app_v2.py
2. Click "Upload AVALANCHE Excel Results" in the sidebar
3. Select your avalanche_results.xlsx file

Adjust citation thresholds in the sidebar:

• Tier 1 (Gold): Default ≥100 citations
• Tier 2 (Silver): Default ≥50 citations
• Tier 3 (Bronze): Default ≥40 citations

Click "🔄 Fetch Citation Network" button. The app will:

• Query OpenAlex API for citation data (~2-3 minutes)
• Build directed citation graph
• Calculate network statistics

• Interact: Hover over nodes, zoom, pan
• Adjust: Change layout algorithm, edge opacity, labels
• Export: Download GraphML or CSV for further analysis

Papers are automatically categorized into visual tiers based on citation impact:

Thresholds are fully customizable via the sidebar.

GraphAvalanche works with any Excel file that has these columns:

• Title: Paper title (string)
• DOI: Digital Object Identifier (string)
• Cited_By: Citation count (integer)

• Year: Publication year (integer) - enables timeline view
• Venue: Journal/conference (string) - shown in hover info
• Abstract: Paper abstract (string) - shown in hover info
• Relevance: Relevance score (float) - from AVALANCHE scoring

Title,Year,DOI,Cited_By,Venue
"Strategies for enzyme prodrug therapy",2017,10.1016/j.addr.201609005,278,Advanced Drug Delivery Reviews
"ADEPT: Trials and tribulations",2012,10.1016/j.bmc.2011年12月02日1,129,Bioorganic Chemistry
"Prodrugs for Targeted Tumor Therapies",2011,10.2174/138161211795428985,100,Current Pharmaceutical Design

• Zoom: Mouse scroll wheel
• Pan: Click and drag
• Hover: View paper details
• Legend: Click to show/hide tiers

• Literature Reviews: Visualize research landscapes
• Gap Analysis: Identify underexplored areas
• Trend Analysis: Track field evolution over time

• Figures: Publication-ready citation networks
• Methods: Document systematic search process
• Discussion: Illustrate research connections

• Course Material: Teach citation analysis
• Student Projects: Guide literature discovery
• Research Training: Demonstrate systematic reviews

• Node Count: Total papers in network
• Edge Count: Citation relationships within dataset
• Network Density: Connectivity measure (0-1)
• Most Cited: Influential papers within network
• Connected Components: Research clusters

# Use in Python with NetworkX
import networkx as nx
G = nx.read_graphml('network.graphml')

# Use in R
library(tidyverse)
data <- read_csv('citation_network_data.csv')

Use the GraphAvalanche modules in your own code:

from graphavalanche import CitationNetwork, MetadataLoader
# Load papers
papers = MetadataLoader.load_excel('avalanche_results.xlsx')
# Build network
network = CitationNetwork(papers)
network.build_connections()
graph = network.build_graph()
# Get statistics
stats = network.get_stats()
print(f"Network has {stats['nodes']} nodes and {stats['edges']} edges")

import networkx as nx
import plotly.graph_objects as go
# Load your graph
G = nx.read_graphml('network.graphml')
# Custom analysis
communities = nx.community.greedy_modularity_communities(G)
betweenness = nx.betweenness_centrality(G)
# Create custom visualization
# ... your code here

# 1. Run AVALANCHE
python avalanche.py 10.1038/nature12373
# 2. Launch GraphAvalanche
streamlit run app_v2.py
# 3. Upload avalanche_results.xlsx
# 4. Click "Fetch Citation Network"
# 5. Export figure for your paper

# For niche fields with lower citation counts:
# 1. Upload your Excel file
# 2. Set Tier 1 = 20, Tier 2 = 10, Tier 3 = 5
# 3. Build network
# 4. Analyze tier distribution

# Compare different literature searches:
# 1. Run AVALANCHE for Topic A → topicA_results.xlsx
# 2. Run AVALANCHE for Topic B → topicB_results.xlsx
# 3. Visualize each in GraphAvalanche
# 4. Compare network structures, densities, tier distributions

GraphAvalanche uses the OpenAlex API (no key required). For faster access:

1. Add your email to get "polite pool" access:

# In app_v2.py, line ~140
headers = {"User-Agent": "GraphAvalanche/1.0 (mailto:your.email@university.edu)"}

2. Rate limiting is built-in (1 second between requests)

Edit tier_colors dictionary in app_v2.py:

tier_colors = {
 'Tier 1': '#FFD700', # Gold
 'Tier 2': '#C0C0C0', # Silver
 'Tier 3': '#CD7F32', # Bronze
 'Other': '#87CEEB' # Sky Blue
}

Contributions are welcome! Areas for improvement:

• Additional layout algorithms (hierarchical, force-atlas)
• Export to other formats (PDF, PNG)
• Clustering analysis (community detection)
• Comparative network analysis (multiple datasets)
• Theme customization (dark mode)
• Performance optimization (large networks >1000 nodes)

See CONTRIBUTING.md for guidelines.

If you use GraphAvalanche in your research, please cite:

GraphAvalanche: Interactive Citation Network Visualizer for AVALANCHE
Harold Mateo Mojica Urrego, University of Navarra-TECNUN, 2026
GitHub: https://github.com/HaroldMate1/GraphAvalanche

For AVALANCHE itself:

AVALANCHE: Automated Federated Literature Discovery Tool
Harold Mateo Mojica Urrego, University of Navarra-TECNUN, 2026
GitHub: https://github.com/HaroldMate1/AVALANCHE

Solution: Ensure your Excel file has columns named exactly Title, DOI, and Cited_By (case-sensitive).

Cause: Papers don't cite each other, or OpenAlex doesn't have citation data.

Solutions:

• Try a different seed paper with more citations
• Check that DOIs are valid
• Verify papers are within the same research domain

Solutions:

• Check internet connection
• Wait a few minutes and retry (rate limits)
• Reduce batch size in code (line ~150: batch_size=25)

Solutions:

• Reduce number of papers (<200 recommended for smooth UI)
• Set edge opacity to 0.2 or lower
• Turn off node labels
• Use "circular" layout (faster than "spring")

MIT License - see LICENSE for details.

• AVALANCHE - Automated literature discovery tool
• Connected Papers - Inspiration for visualization style
• Gephi - Advanced network visualization
• VOSviewer - Bibliometric network visualization
• OpenAlex - Open citation database

• Author: Harold Mateo Mojica Urrego
• Institution: University of Navarra-TECNUN
• GitHub: @HaroldMate1
• Issues: GitHub Issues

• OpenAlex for providing open citation data
• Plotly for interactive visualization capabilities
• NetworkX for graph algorithms
• Streamlit for the web framework
• Connected Papers for visualization inspiration

Made with ❤️ for the research community

Transform your literature reviews into interactive visual insights 🚀