PRISTINE: Profile-based Inference for Statistical Networks

PRISTINE is a modular framework for phylogenetic likelihood inference, built in PyTorch ≥ 2.6 and Python ≥ 3.10.

It supports: - Efficient likelihood computation with Felsenstein's pruning algorithm - Model-based simulation and inference (e.g., GTR, BDS) - Parameter profiling and Laplace approximation - Relaxed and strict molecular clocks - Transparent and flexible parameter management

---

📌 Quick Start

1. Install locally:

pip install -e .

1. Run a minimal inference:

   from pristine import optimize, binarytree, gtr
   ...
   

2. See tutorials and covered scientific questions

---

📦 Components

Module	Description
felsenstein.py	FPA-based log-likelihood engine
gtr.py	GTR substitution model
sequence.py	Sequence simulation and pattern collapse
molecularclock.py	Strict and relaxed clock models
laplace_estimator.py	Laplace approximation of uncertainty
likelihood_profiler.py	Confidence intervals via profiling
binarytree.py	Tree structure and simulation
edgelist.py	Time-aware representation of trees