InSoLiTo

InSoLiTo (Interactions of Software for Life Sciences Tools) shows which research software tools are used together, based on how often their papers are cited together (co-citation). Live version: https://test-insolito.openebench.bsc.es/

This is the repository of BioHackathon Europe 2026, Project 17: Bridging Literature and Practice: Tool Co-citation Networks from Publications and Registered Workflows. On top of co-citation, it adds directed edges: which tool is cited before which in the text of an article.

Repository layout

frontend/              Web app (Nuxt 4, Vue 3, Cytoscape.js)
DB/                    InSoLiTo database
├── pipeline/          Scripts that build it: PubMed + tool catalogue → SQLite → CSV → Neo4j
├── raw_data/          Inputs and intermediate files (not in git)
├── CONF/ DATA/ LOGS/ PLUGINS/ InSoLiToImport/   Neo4j volumes
└── *.json             Static data bundled into the web app
directed-graph/        Directed edges from citation order
├── europepmc/         Download and parse the full-text corpus from Europe PMC
├── biotools/          Publications of each bio.tools tool
└── citation-order/    Build the directed edges (steps 01 → 05b)
deploy/                Apache config, and the web app config for a local Neo4j
docker-compose.yml     Web app build + Neo4j + Apache
requirements.txt       Python packages for DB/pipeline/ and directed-graph/citation-order/

Run locally with Docker

You need Docker with Compose v2.

cp .env.example .env
mkdir -p DB/InSoLiToImport frontend/.output/public
docker compose up -d

• On Linux, set HOST_UID and HOST_GID in .env to your id -u and id -g, so the files the containers create belong to you.
• The first run builds the web app inside Docker (a few minutes). Then open http://localhost:8080 (INSOLITO_PORT). The Neo4j browser is at http://localhost:7474 (NEO4J_HTTP_PORT).
• By default the web app queries the public InSoLiTo database. To use your local Neo4j, set INSOLITO_CONFIG=./deploy/neo4j-local.json in .env and run docker compose up -d --force-recreate.

To work on the web app only, without Docker (Node ≥ 22.19):

cd frontend
npm install
npm run dev

Populate the local Neo4j

Create the Python environment once, from the repository root:

python3 -m venv .venv
.venv/bin/pip install -r requirements.txt

The import deletes everything in the Neo4j database it points to. Use it only against your local Neo4j. The connection is set in DB/pipeline/neo4j/*.ini (url_neo4j_server): its port must match NEO4J_BOLT_PORT in your .env.

Option A: published data (fast)

The data behind the current InSoLiTo (2023) is on Zenodo: 10.5281/zenodo.7524268. With Neo4j running (docker compose up -d), from the repository root:

wget -P DB https://zenodo.org/records/7524268/files/InSoLiToImport.zip
cd DB && unzip -o InSoLiToImport.zip && cd ..
.venv/bin/python DB/pipeline/neo4j/CreateNeo4jDataset.py DB/pipeline/neo4j/local-config.ini

To see it in the web app, set INSOLITO_CONFIG=./deploy/neo4j-local.json in .env and run docker compose up -d --force-recreate.

Option B: rebuild from scratch (hours)

Put the inputs in DB/raw_data/:

• pubmed_baseline/: the PubMed baseline XML files (~54 GB), from https://ftp.ncbi.nlm.nih.gov/pubmed/baseline/
• oeb_tools.json: the OpenEBench tool catalogue, from https://openebench.bsc.es/monitor/rest/search?=publications

Stop Neo4j (docker compose down) and run, from the repository root:

nohup DB/pipeline/run_pipeline.sh > DB/raw_data/pipeline.log 2>&1 &

It builds the SQLite database, exports the CSV files to DB/InSoLiToImport/, starts Neo4j, imports them and writes the web app JSON files to DB/raw_data/json_full/ (copy them to DB/ to use them in the web app). REUSE_PUBLICATIONS=1 reuses the publications table of a previous run.

Directed graph

Inputs:

• Corpus: the Parquet tables in directed-graph/europepmc/data/export/. To build them, see directed-graph/europepmc/README.md (those scripts run from their own folder, with their own requirements.txt).
• Tool publications: directed-graph/biotools/data/biotools_publications.csv (in git; refresh it with pull_biotools.py).

Run each step from the repository root, in order (seconds to a few minutes each). Each one writes its result to directed-graph/citation-order/data/, and the next step reads it:

.venv/bin/python directed-graph/citation-order/01_match_tools.py

Step	What it does
01_match_tools.py	Finds which bio.tools papers are cited in the corpus (by DOI, PMID or PMCID)
01b_top_works.py	Check: the most cited works and their tool, if any
02_filter_articles.py	Keeps research articles
03_order_tools.py	Position of each tool in each article (Methods section by default, SECTION at the top)
04_directed_pairs.py	One row per tool pair and article, with the earlier cited tool as source
05_sum_pairs.py	One row per pair and direction: number of articles, share and distance
05b_cytoscape.py	Filtered edge list for Cytoscape (thresholds at the top of the script)

To read how two tools are cited together, pair_contexts.py takes two bio.tools IDs. It counts the articles that cite both, grouped by whether the two citations share a sentence or a paragraph, and by which tool comes first. It reads the delta and tool gap from the outputs of steps 3 to 5 in directed-graph/edges-processed/ (EDGES at the top of the script). It writes directed-graph/citation-order/data/pair_contexts/<tool_a>__<tool_b>.html. The page shows the pair's two edges from directed_edges.csv, and histograms of both distances over the pair's articles. For 15 articles, split across the groups by their size, it downloads the full text from Europe PMC and shows where each tool falls among the references, and the paragraphs that cite either tool with each tool's citations highlighted in its own colour:

.venv/bin/python directed-graph/citation-order/pair_contexts.py star deseq2

Data sources

• PubMed baseline (NLM): reference lists, for co-citation
• OpenEBench: tool catalogue
• EDAM ontology: topics and operations
• Europe PMC: Open Access full text, for citation order
• bio.tools: tool publications

Supporting files of the current InSoLiTo: 10.5281/zenodo.13740843.

License

MIT. See LICENSE.txt.