SimPhyNI

Overview

SimPhyNI (Simulation-based Phylogenetic iNteraction Inference) is a phylogenetically-aware framework for detecting evolutionary associations between binary traits (e.g., gene presence/absence, major/minor alleles, binary phenotypes) on microbial phylogenetic trees. This tool leverages phylogenetic information to correct for spurious associations caused by the relatedness of sister taxa.

This pipeline is designed to:

• Infer evolutionary parameters for traits (gain/loss rates, time to emergence, ancestral states)
• Estimate trait co-occurrence null models through independent simulation of traits
• Output statistical results for associations

Graphical Workflow:

---

Getting Started

Installation

First, ensure bioconda and conda-forge channels are configured:

conda config --add channels conda-forge
conda config --add channels bioconda

Create a new environment:

conda create -n simphyni
conda activate simphyni

Then install SimPhyNI from bioconda:

conda install simphyni

Test installation:

simphyni version

Input Specifications

1. Phylogenetic Tree (.nwk)

• Standard Newick format.
• Must be rooted (both outgroup and midpoint are acceptable).
• Tip labels must match the sample name index in your traits file.
• Branch lengths are required for accurate rate estimation.

2. Traits File (.csv or .parquet)

• Rows: Genomes/Samples (matching tree tips).
• Columns: Binary traits (0 = Absent, 1 = Present; non-numerical values will be set to 1 and blank values will be set to 0).
• Header: Required (Trait names).
• Index: The first column must contain sample names matching tree tip labels.

For large datasets (>10,000 traits), Parquet format is strongly recommended. Parquet enables column-projection and row-predicate pushdown, so only the data actually needed is loaded from disk, dramatically reducing peak memory use.

Example traits.csv:

Sample,PhenotypeX,GeneA,GeneB
E_coli_1,1,0,1
E_coli_2,1,1,0
E_coli_3,0,0,1

Starting from FASTA files and looking for gene-gene or gene-phenotype associations?

If you have raw genome assemblies (FASTA) and need to generate the necessary inputs (gene presence/absence and a phylogenetic tree), we provide a dedicated pipeline: SimPhyNI-Prelude.

This Snakemake workflow is configured for HPC and automates the following steps:

• Annotation (Prokka)
• Pangenome Analysis (Panaroo)
• Tree Construction (PopPUNK or RAxML)
• Formatting (Preparation for SimPhyNI)
• SimPhyNI Analysis (This repository)

Any steps may be bypassed by providing existing data (e.g. gene annotations, phylogenetic tree).

For those familiar with Snakemake, rules can be edited, added, or removed to suit your needs.

---

Usage

Run mode (single-run)

simphyni run \
  --sample-name my_sample \
  --tree path/to/tree.nwk \
  --traits path/to/traits.csv \
  --run-traits 0,1,2 \
  --outdir my_analysis \
  --cores 4 \
  --temp-dir ./tmp \
  --min_prev 0.05 \
  --max_prev 0.95 \
  --plot

Key flags:

• --run-traits — Comma-separated list of N values (e.g. 0,1,2) to designate the first N traits in the file as query traits for a one-vs-rest comparison. Use ALL (default) for all-against-all. For example, --run-traits 0,1,2 tests the first 3 traits against all others.
• --include-flagged — By default, traits whose Poisson null distributions are detected as miscalibrated (sparse events combined with highly asymmetric eligible regions) are excluded from testing and assigned p=0.5. Pass --include-flagged to simulate these traits anyway; their results will be marked with null_calibrated=False in the output so you can assess them separately. Access the list of flagged traits via Sim.get_flagged_traits() when using the Python API.
• --min_prev / --max_prev — Minimum and maximum tip prevalence thresholds for traits to be included (default: 0.05 / 0.95).
• --plot — Generate heatmap summaries of results.
• --save-object — Save the full analysis object as a .pkl file (not recommended for analyses > 1,000,000 trait pairs).

Run mode (batch)

Create a samples.csv file:

Sample,Tree,Traits,run_traits,MinPrev,MaxPrev
run1,tree1.nwk,traits1.csv,ALL,0.05,0.95
run2,tree2.nwk,traits2.csv,"0,1,2",0.05,0.90

• run_traits, MinPrev, and MaxPrev are optional columns that will use default values if not provided.
• run_traits must be ALL (case-sensitive) or a comma-separated list of N values (e.g. 0,1,2 to query the first 3 traits against all others).

Then execute:

simphyni run --samples samples.csv --cores 16

Run with HPC

First, download the cluster profile template:

simphyni download-cluster-profile

Edit cluster_profile/config.yaml for your computing cluster, then install the appropriate Snakemake executor from the available catalog: https://snakemake.github.io/snakemake-plugin-catalog/index.html (SLURM shown below):

pip install snakemake-executor-plugin-slurm

Run SimPhyNI with the --profile flag:

simphyni run --samples samples.csv --profile cluster_profile

For all run options:

simphyni run --help

Example data

Download and run example inputs using:

simphyni download-examples
simphyni run --samples example_inputs/simphyni_sample_info.csv --cores 8 --plot

---

Outputs

Outputs for each sample are placed in structured folders in the working directory or specified output directory in subdirectories by sample name, including:

Main Result Files

simphyni_results.csv Contains the statistical results for all tested trait pairs.

Column	Description
T1 / T2	Identifiers for the two traits being compared.
direction	Direction of association: 1 = Positive, -1 = Negative.
effect size	Variance-adjusted magnitude of the association.
prevalence_T1 / prevalence_T2	Fraction of samples containing each trait (0.0 to 1.0).
pval_naive	Raw empirical P-value from the simulation.
pval_bh	Benjamini-Hochberg FDR correction (recommended for phenotype-genotype tests).
pval_by	Benjamini-Yekutieli FDR correction (recommended for genotype-genotype tests; accounts for correlated hypotheses, which is the typical case for phylogenetically structured data).
pval_bonf	Bonferroni correction (strictest; use when a small number of specific hypotheses are tested).
null_calibrated	True for standard results. False for pairs involving traits whose null distributions were detected as miscalibrated (only present when --include-flagged is used).

Additional Outputs

• simphyni_results.csv: Main results table (see above).
• simphyni_object.pkl: Optional file containing the completed analysis object, parsable with an active SimPhyNI environment. Controlled with the --save-object flag (not recommended for large analyses > 1,000,000 comparisons).
• Plots: Heatmap summaries of tested associations (if --plot is enabled).

---

Directory Structure

SimPhyNI/
├── simphyni/               # Core package
│   ├── Simulation/         # Simulation engine
│   ├── scripts/            # Pipeline scripts
│   ├── Snakefile.py        # Workflow definition
│   ├── simphyni_cli.py     # Command-line entry points
│   └── envs/simphyni.yaml  # Conda environment (used by Snakemake)
├── tests/                  # Testing suite
├── conda-recipe/           # Build recipe
├── cluster_profile/        # Cluster config template for HPC
├── example_inputs/         # Example inputs to run SimPhyNI
└── pyproject.toml

---

Contact

For questions, please open an issue or contact Ishaq Balogun at https://github.com/jpeyemi.

Citation

If you use SimPhyNI in your research, please cite:

High Precision Binary Trait Association on Phylogenetic Trees Ishaq O Balogun, Christopher P Mancuso, Tami D Lieberman bioRxiv 2025.12.24.696407; doi: https://doi.org/10.64898/2025.12.24.696407