PhyloNaP Pipeline

Automated construction of the PhyloNaP enzyme phylogeny database — a comprehensive resource linking natural product biosynthetic enzymes (tailoring enzymes) with phylogenetic, functional, and genomic context.

Overview

The pipeline processes protein sequences from multiple annotation sources (MITE, MIBiG, SwissProt, AntiSMASH), clusters them, builds phylogenies, and produces a searchable database with superfamily-grouped datasets.

Input FASTA files
      │
      ▼
┌─────────────────────┐
│ 01. MMseqs2 cluster │
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 02. Filter clusters │  (>10 seqs, ≥3 BGC-linked)
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 03. MAFFT + trimAl  │
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 04. Fast pre-trees  │
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 05. TreeCluster     │  Clade-based subclustering
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 06. Split + re-QC   │  Subclusters → align → trim → AliStat
└────────┬────────────┘
         ▼
┌─────────────────────────────┐
│ 07. Cascade filtration      │  Cr ≥ 0.6, Ca > 0.7,
│     (remove bad sequences,  │  retention ≥ 0.75,
│      re-align, quality gate)│  mean length > 70, n ≥ 10
└────────┬────────────────────┘
         ▼
┌─────────────────────┐
│ 08. Annotate        │  MITE > MIBiG > SwissProt > AS_db
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 09. Deduplicate     │  Remove identical sequences
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 10. Final filter    │  ≥1 BGC-linked sequence required
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 11. Build trees     │  FastTree
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 12. Organize        │  PT IDs, merged FASTA
└────────┬────────────┘
         ▼
┌─────────────────────┐
│ 13. Root trees      │  Taxonomy outgroup + MAD fallback
└────────┬────────────┘
         ▼
┌──────────────────────────┐
│ 14. Build database       │  JSON + SQLite, superfamily grouping
└────────┬─────────────────┘
         ▼
┌──────────────────────────┐
│ 15. Reference DB         │  MMseqs2 dereplicated search DB
└──────────────────────────┘

Quick Start

Prerequisites

Platform: macOS or Linux only (Windows is not supported).

1. Create the Conda environment

conda env create -f environment.yml
conda activate phylonap_createdb

2. Install AliStat

AliStat is required but not available via conda. Use the provided helper script:

./install_alistat.sh            # downloads, builds, and installs to ./tools/alistat

Or, if you already have AliStat installed, set the path in config/config.yaml:

tools:
  alistat: "/path/to/your/alistat"

To build manually:

wget https://github.com/thomaskf/AliStat/archive/refs/heads/master.tar.gz -O AliStat.tar.gz
tar -zxvf AliStat.tar.gz
cd AliStat-master
make
# copy the resulting 'alistat' binary somewhere on your $PATH

3. Configure

Edit config/config.yaml to set your input data paths and parameters.

Required input files:

• FASTA sequences from AntiSMASH (AS_db), MIBiG, SwissProt, and/or MITE
• Annotation tables (TSV) for each data source
• ID mapping file linking cluster IDs to original accessions

3. Run

# Full pipeline
snakemake --cores 8

# Dry run (preview steps)
snakemake --cores 1 -n

# Run up to a specific step
snakemake --cores 8 --until annotate

# Run a single step
snakemake --cores 4 build_trees

Configuration

All parameters are centralized in config/config.yaml:

Section	Description
output_dir	Base output directory
input_data	Paths to FASTA input files
annotations	Paths to annotation/metadata TSVs
id_mapping	Sequence ID mapping file
tools	Paths to external tools (mmseqs, mafft, etc.)
params	All pipeline thresholds and parameters

Key Parameters

Parameter	Default	Description
mmseqs_sensitivity	7.5	MMseqs2 clustering sensitivity
mmseqs_min_aln_len	80	Minimum alignment length for clustering
min_cluster_size	10	Minimum sequences per cluster
min_bgc_sequences	3	Minimum BGC-derived sequences per cluster
treecluster_threshold	4.0	TreeCluster clade distance threshold
min_cr	0.6	Minimum per-column residue score
min_ca	0.7	Minimum alignment column score
min_retention	0.75	Minimum trimAl retention fraction
min_mean_length	70	Minimum mean sequence length after trimming
fasttree_model	-lg -gamma	FastTree substitution model

Output Structure

results/
├── 01_cluster_results/     MMseqs2 clustering output
├── 02_filtered_clusters/   Per-cluster FASTA files
├── 03_trimmed_alignments/  MAFFT + trimAl output
├── 04_fast_trees/          Pre-trees for subclustering
├── 05_treecluster/         TreeCluster assignments
├── 06_subclusters/         Split subclusters + QC stats
├── 07_filtered_datasets/   Quality-filtered datasets
├── 08_annotations/         Merged annotation tables
├── 09_deduplicated_*/      Deduplicated alignments + annotations
├── 10_final_*/             Final filtered datasets
├── 11_trees/               ML phylogenies
├── 12_datasets/            Organized PT-numbered folders
│   ├── PT000001/
│   │   ├── PT000001.fasta
│   │   ├── PT000001_al.fa
│   │   ├── PT000001.nwk
│   │   └── PT000001_annotations.tsv
│   ├── pt_id_mapping.tsv
│   └── all_sequences_merged.fasta
├── 13_rooting_summary.tsv
├── 14_database/
│   ├── db_structure.json
│   └── phylonap.db
└── 15_reference_db/
    ├── dereplicated.fasta
    └── mmseqs_db/

Dependencies

Managed via Conda (environment.yml):

Bioinformatics tools:

• MMseqs2 (sequence clustering + reference DB)
• MAFFT (multiple sequence alignment)
• trimAl (alignment trimming)
• FastTree (phylogenetic tree inference)
• AliStat (alignment quality statistics)
• TreeCluster (clade-based tree cutting)

Python packages:

• BioPython, pandas, numpy, PyYAML

R packages (for MAD rooting):

• ape, phangorn, Rcpp, inline

Annotation Priority

Sequences are annotated using a priority cascade:

1. MITE — Experimentally characterized enzyme reactions
2. MIBiG — Validated BGC-associated enzymes
3. SwissProt — Curated protein annotations
4. AS_db — AntiSMASH-predicted BGC enzymes

When a sequence appears in multiple sources, higher-priority annotations take precedence. MITE_IDs are assigned only by direct sequence match (not propagated across clusters).

Tree Rooting Algorithm

Trees are rooted using a bipartition-based taxonomy outgroup detection:

1. All internal edges are evaluated as potential rooting positions
2. Both sides of each edge are tested at each taxonomy level (Superkingdom → phylum → class → order → Genus → Species)
3. Mode A (pure outgroup): minority taxon forms a pure clade
4. Mode B (dominant monophyly): rooting makes the dominant taxon monophyletic
5. Guards: annotation coverage thresholds, stem length checks, single-genus species protection
6. Fallback: MAD (Minimal Ancestor Deviation) rooting via R

Cascade Filtration

The quality filtration preserves phylogenetic signal through iterative refinement:

1. Remove bad sequences: column residue score (Cr) < 0.6
2. Re-align remaining sequences (MAFFT)
3. Re-trim (trimAl automated1)
4. Quality gate: Ca > 0.7, retention ≥ 0.75, mean_length > 70, n ≥ 10

Datasets failing the quality gate are excluded.

Citation

If you use PhyloNaP in your research, please cite:

[Citation to be added upon publication]

License

[License to be specified]