StructPhylogeny

StructPhylogeny is a Snakemake pipeline for building parallel structure-based and sequence-based phylogenies from a folder of protein structures.

It is designed around the workflow direction of Lakshmi et al. (2015), but implemented with modern tooling:

• GTalign for all-vs-all 3D structural alignment and comparison
• an SDM-like structural dissimilarity matrix computed from pairwise alignments
• a structure tree inferred from the SDM matrix
• sequence extraction directly from the input structures
• MAFFT for multiple sequence alignment
• IQ-TREE 3 for maximum-likelihood sequence phylogeny
• summary QC, distance-matrix comparison, and heatmap/report outputs

Disclaimer

• This package requires CUDA - it would be possible to run this without cuda but a different installation of GTalign would be needed - modify the mamba env based on the GTalign repository docs to use the CPU version
• This package was used and developed as part of my diploma thesis; no official publication regarding this exists as of now. If needed, cite the tools used here, such as GTalign, MAFFT, IQTREE3, etc.

Environment

The repository includes a mamba environment specification in environment.yml.

Create and activate it with:

mamba env create -f environment.yml
mamba activate structphylogeny

Input

Place protein structures in a directory such as data/my_structures/.

Supported structure formats:

• .cif
• .mmcif
• .pdb
• .ent

The current default dataset is:

data/mouse_lipocalin_structures

Run

The default configuration is in config/config.yaml.

Run the full pipeline with:

snakemake --cores 8 # For example data
snakemake --cores 8 --configfile path/to/your/config.yaml # For your defined folder of proteins and flags for MAFFT or IQTREE

Or dry-run it first:

snakemake --cores 8 --dry-run

Outputs

Main outputs land under results/:

• results/manifest/structures.tsv: input manifest and basic QC
• results/sequences/proteins.fasta: extracted sequences
• results/sequences/proteins.aligned.fasta: MAFFT alignment
• results/trees/sequence.treefile: IQ-TREE sequence phylogeny
• results/structure/gtalign_pairwise.tsv: parsed GTalign pairwise metrics
• results/structure/sdm.tsv: structural dissimilarity matrix
• results/trees/structure_sdm.nwk: structure tree from the SDM matrix
• results/reports/sdm_heatmap.png: SDM heatmap
• results/reports/sequence_distance_heatmap.png: sequence p-distance heatmap
• results/reports/matrix_correlation.tsv: agreement between structure and sequence distances
• results/reports/report.md: concise summary report

Notes

• The paper used DALI and a PHYLIP distance-tree method. This repository uses GTalign for structural comparison, as it more modern, faster and better tool.
• The SDM implementation follows the Lakshmi et al. formula and computes PFTE from topologically equivalent residues as GTalign reports them.
• IQ-TREE 3 is used for the sequence phylogeny. The structure phylogeny is inferred from the SDM distance matrix with a neighbor-joining distance-tree method.
• By default, if a structure has multiple chains, the pipeline uses the longest protein chain.

Tests

Run the lightweight unit tests with:

PYTHONPATH=src pytest