GFViewer

Visualize the localization of multigene families across the chromosomes of a genome, and quantify how those families are distributed.

GFViewer draws round-capped chromosome ideograms with each family in its own colour (+ strand above the axis, - below), optionally marks centromeres, and can compute localization statistics. Output is fully vector and can be saved as PDF, SVG, PNG, JPG, TIFF or EPS.

Version 2 rewrites the rendering engine on top of matplotlib (the old BioPython/BasicChromosome engine is kept at gfviewer.legacy for one release), removes the 19-family ceiling, adds BED / GFF3 / GTF input, a style system, an analytics module, and an interactive web portal.

Install

pip install gfviewer                      # the `gfviewer` command-line tool

From a source checkout (development, tests, or the web portal):

conda env create -f environment.yml      # or: python -m venv .venv && source .venv/bin/activate
conda activate gfviewer
pip install -e ".[web,dev]"               # editable, with Flask + pytest

A full pip-installation and command-line walkthrough — isolated environments, every flag with worked examples, troubleshooting — is in docs/INSTALL.md, and is mirrored on the web portal's Install page. Build distributable artifacts with scripts/build_wheel.sh.

Command line

gfviewer -d genes.tsv -g genome.fasta -o out/ \
         -f pdf svg png --analytics -cen \
         --tick-style lollipop --title "My organism"

Key options (gfviewer --help for the full list):

Input formats

• Table — columns gene_id, gene_family, chromosome, start, end, strand (1-based, inclusive; header synonyms accepted). A centromere value in gene_family (strand 0) marks a centromere.
• BED — 0-based half-open, converted automatically. One file → column 4 is the family; several files → each file's base name is the family; one file plus a mapping file → family from the map.
• GFF3 / GTF — 1-based; gene-like feature types kept; family taken from --family-attr (default search: gene_family, Family, gene_biotype, Name, …).

Analytics

--analytics writes analytics_*.csv / .bed, analytics_summary.json and four figures (analytics_genes_per_family, analytics_positional_profile, analytics_ripley, analytics_family_proximity — one per requested figure format). It computes:

• per-family counts split into genes on chromosomes vs. unplaced/stray contigs, linear density, gene length, strand fraction;
• permutation tests for telomere- and (with a centromere track) centromere-proximal bias, plus p-arm / q-arm occupancy;
• tandem-array detection and a tandem / proximal / dispersed duplication mode per family;
• multi-scale clustering — edge-corrected 1-D Ripley's K/L with a permutation envelope (--ripley-scales);
• chromosome-enrichment (binomial, per family × chromosome), strand bias, and per-chromosome family diversity (Shannon / evenness);
• a binned positional density profile ("metachromosome" plot);
• a family × family proximity matrix with average-linkage clustering;
• multigene-family hotspots — a Poisson window scan, merged and written as a table and a BED file (--hotspot-window);
• optional pairwise co-localization (--colocalization).

Every per-family / per-pair / per-window test carries a Benjamini–Hochberg q_value. See the Downloadable outputs section of the Help page for the column-by-column layout of each file.

Web portal

# development
python flaskapp.py                       # http://localhost:5001

# production
gunicorn -w 1 --threads 4 -b 0.0.0.0:5001 "gfviewer_web:create_app()"

Uploads are rendered on a background thread pool behind an async job API (POST /api/jobs → GET /api/jobs/<id>/status → results page). The results page embeds the SVG with an editor for choosing which families and chromosomes to draw, recolouring, moving the legend and labels, changing fonts and mark style, toggling whether titles are baked into the image files, then re-rendering and exporting. Analytics figures are shown inline, and two buttons download everything as a ZIP — either as produced, or re-rendered in every figure format. GET /api/health is a readiness probe.

Usage monitor. A privacy-respecting counter (/stats, or JSON at /api/stats) tracks page views, unique visitors per day (a salted daily hash — no IPs, cookies or other personal data are stored), jobs submitted / completed / failed, example-dataset runs and downloads, with a rolling ~120-day daily series. Counters persist to instance/usage.json.

Configuration (environment variables): GFVIEWER_DATA_DIR, GFVIEWER_MAX_UPLOAD_MB (25), GFVIEWER_WORKERS (2), GFVIEWER_JOB_TTL_HOURS (24), GFVIEWER_USAGE_FILE, GFVIEWER_STATS_TOKEN (require ?token= on /stats), SECRET_KEY, PORT.

Example datasets

python tests/make_fixtures.py builds every bundled dataset into static/tests/: the three Babesia sets, the 6-family set re-expressed in every input format (formats/ — BED, per-family BED, BED + mapping, GFF3, GTF), and two synthetic sets — 10 random gene families on the Arabidopsis (TAIR10) chromosomes and 20 on C. elegans (WBcel235). The web home page lists them all with Run (submits the job) and Download buttons, plus Download every dataset (ZIP).

Tests

python tests/make_fixtures.py            # once, to build the example datasets
pytest -q

Citation

Chakravarty S. & Lonardi S. Visualizing the localization of multigene families with GFViewer. Development supported by NIH grant 1-R01-AI169543-01.