consite 0.1.4

ConSite: conserved-domain alignment and conserved-site visualization from protein FASTA

ConSite

ConSite takes a protein FASTA sequence, finds conserved domains via local Pfam/HMMER, aligns your sequence to each domain, scores per-position conservation, and renders publication-quality figures and structured outputs.

Features

• FASTA → Pfam domain search (local HMMER)
• Automatic alignment to each hit using the family's Pfam SEED HMM
• Per-position conservation: entropy, Jensen–Shannon divergence (JSD), consensus frequency, coverage
• Conserved site calling (top-X% by JSD, per domain)
• Publication-quality visualization
  • Linear domain map with hollow-red conserved sites
  • MSA gradient panels (from Pfam SEED): brightness-floored grayscale so letters remain legible; supports species labels, optional query row at top, gap glyphs (dash/dot/none), coverage masking/weighting
  • Per-domain alignment panels for the query with optional conservation background scale
• Reproducible outputs (JSON, TSV, PNG, Stockholm) and clear CLI logging

Installation

Option A: PyPI (recommended)

python -m pip install consite

Prerequisites

• Python ≥ 3.10
• HMMER 3.x in your PATH
• Pfam database files (see Quick Start)

Install HMMER:

• macOS (Homebrew): brew install hmmer
• Linux (APT): sudo apt-get update && sudo apt-get install hmmer
• Windows (conda): conda install -c conda-forge hmmer

Verify: hmmsearch --version

Quick Start

1) Get the Pfam database

Use the helper script:

chmod +x scripts/*.sh
./scripts/get_pfam.sh

This downloads Pfam-A.hmm and Pfam-A.seed, uncompresses, and runs hmmpress.

(Manual alternative is in "Manual Setup" below.)

2) Run ConSite (demo)

consite \
  --fasta examples/P05362.fasta \
  --pfam-hmm pfam_db/Pfam-A.hmm \
  --pfam-seed pfam_db/Pfam-A.seed \
  --out results \
  --id P05362

Manual Setup (from source)

git clone https://github.com/yangli-evo/ConSite.git
cd ConSite
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -e .

# (then run ./scripts/get_pfam.sh, or manually download + hmmpress)

Manual Pfam download:

mkdir -p pfam_db
curl -L -o pfam_db/Pfam-A.hmm.gz https://ftp.ebi.ac.uk/pub/databases/Pfam/current_release/Pfam-A.hmm.gz
curl -L -o pfam_db/Pfam-A.seed.gz https://ftp.ebi.ac.uk/pub/databases/Pfam/current_release/Pfam-A.seed.gz
gunzip -f pfam_db/Pfam-A.hmm.gz pfam_db/Pfam-A.seed.gz
hmmpress pfam_db/Pfam-A.hmm

Usage

Basic:

consite \
  --fasta myprotein.fasta \
  --pfam-hmm pfam_db/Pfam-A.hmm \
  --pfam-seed pfam_db/Pfam-A.seed \
  --out results \
  --topn 5 \
  --cpu 8 \
  --jsd-top-percent 15 \
  --log results/run.log

MSA panel tuned example (SEED gradient = JSD, labels with species+id, include query row, safe brightness):

consite \
  --fasta examples/GS2.fasta \
  --pfam-hmm pfam_db/Pfam-A.hmm \
  --pfam-seed pfam_db/Pfam-A.seed \
  --out results \
  --id GS2 \
  --msa-panel-nseq 8 \
  --msa-panel-metric jsd \
  --msa-labels species+id \
  --msa-include-query \
  --msa-min-brightness 0.28 \
  --panel-min-brightness 0.22

Outputs

For each run you'll get a folder results/<id>/ containing:

• query.fasta – input sequence
• hits.json – Pfam hits (family, coords, scores)
• scores.tsv – per-position tracks (columns: pos, in_domain, jsd, entropy, is_conserved)
• domain_map.png – linear domain map with conserved sites
• *_panel.png – per-domain query panels with conserved sites (hollow red)
• *_msa.png – Pfam SEED MSA panels with grayscale conservation gradient, labels, and optional query row
• *_sim.png – pairwise % identity heatmap among panel sequences (RF-masked columns)
• *_sim.tsv – pairwise % identity matrix (TSV)
• *_aligned.sto – Stockholm alignment of query to each family HMM
• hmmsearch.domtblout – raw HMMER domain table
• run.log – external tool logs

Command-line Options

Option	Description	Default
--fasta	input protein FASTA	Required
--pfam-hmm	Path to Pfam-A.hmm (pressed)	Required
--pfam-seed	Path to Pfam-A.seed	Required
--out	Output directory	Required
--id	Custom run ID (subfolder name)	FASTA header
--topn	Number of top Pfam hits to analyze	2
--cpu	HMMER threads	4
--jsd-top-percent	Top % (by JSD) called conserved within each domain	10.0
--log	Append external tool output to this file	results/<id>/run.log
--quiet	Suppress console output	False
--keep	Keep existing output folder (don't overwrite)	False

MSA panel (SEED) controls

Option	Description	Default
--msa-panel-nseq	Rows to display from the SEED alignment	8
--msa-panel-metric	Gradient metric: entropy → uses 1-entropy, or jsd	entropy
--msa-labels	Row labels: id, species, species+id	species+id
--msa-include-query	Prepend query row to the MSA panel	off
--msa-min-brightness	Floor for background brightness (keeps letters legible)	0.26
--msa-min-coverage	Mask columns below this coverage in panels (0–1)	0.30
--cons-weight-coverage	Weight conservation by coverage (alpha)	1.0
--mask-inserts	Use RF to mask insert columns	True
--gap-glyph	Gap rendering in MSA: dash, dot, or none	dash
--gap-cell-brightness	Brightness used in gap cells	0.90
--write-sim-matrix / --no-write-sim-matrix	Write pairwise % identity matrices for MSA panels	True

Per-domain panel (query) controls

Option	Description	Default
--panel-min-brightness	Brightness floor for query panels	0.18

How It Works

1. Domain detection: hmmsearch against Pfam-A HMM library (GA thresholds).
2. SEED extraction: the matched family's block is pulled from Pfam-A.seed.
3. Model building: hmmbuild produces a family HMM.
4. Query alignment: hmmalign aligns your sequence to the family model.
5. Scoring:
   • MSA panels use SEED-based per-column scores (JSD or 1-entropy), with optional coverage masking/weighting.
   • Conserved sites are called on the query alignment per domain (top-X% by JSD).
6. Visualization: domain map, query panels, and SEED MSA gradient panels.

Example (ICAM1)

For examples/P05362.fasta, ConSite typically finds:

• PF03921 (~25–115): Ig-like domain
• PF21146 (~219–308): Ig-like domain

You'll see two *_panel.png, two *_msa.png, a domain_map.png, plus hits.json and scores.tsv.

Troubleshooting

• command not found: hmmsearch Install HMMER and ensure it's on your PATH ( which hmmsearch ).

• No such file or directory: pfam_db/Pfam-A.hmm Run ./scripts/get_pfam.sh or download manually and hmmpress.

• Large/verbose logs Use --quiet and/or inspect run.log.

Development

ConSite/
├── src/consite/
│   ├── cli.py            # CLI
│   ├── hmmer_local.py    # HMMER wrappers
│   ├── parse_domtbl.py   # HMMER output parsing
│   ├── pfam.py           # Pfam SEED extraction
│   ├── msa_io.py         # Stockholm I/O (+RF)
│   ├── conserve.py       # Scoring (entropy/JSD/coverage)
│   └── viz.py            # Plots (domain map, panels, MSA gradient)
├── scripts/              # Helper scripts (get_pfam, quickstart)
├── examples/             # Example FASTA files
└── pfam_db/              # Pfam files (user-provided)

Dependencies: biopython ≥1.81, numpy ≥2.0, pandas ≥2.0, scipy ≥1.16, matplotlib ≥3.7, Python ≥3.10.

Notes & Roadmap

• Remote CDD mode is stubbed (local Pfam/HMMER path is fully supported).
• Current conserved-site calls are relative (top-X%); absolute thresholds are planned.

Citation

Joey Wagner, Yang Li. ConSite: conserved-domain alignment and conserved-site visualization from protein FASTA.

License

MIT (see LICENSE).