Monte-Carlo generation of synthetic multiple sequence alignments along phylogenetic trees using a protein language model

Lab Immersion at EPFL
Lab: Bitbol Lab – Laboratory of Computational Biology and Theoretical Biophysics
Professor: Anne-Florence Bitbol
Supervisors: Umberto Lupo, Damiano Sgarbossa, Cyril Antoine Malbranke

Table of Contents

1. Description
   • Built With
2. Getting Started
   • Prerequisites
   • Installation
3. Usage
4. License

Description

This project generates a phylogenetic tree from a natural multiple sequence alignment (MSA) using either FastTree or IQTree. From this tree, it produces a synthetic MSA through a Metropolis–Hastings algorithm for Markov Chain Monte Carlo (MCMC) employing the probabilities given by the ESM2 model. The aim is to acquire synthetic data to fine-tune the MSA transformer.

(back to top)

Built With

• 
• 
• 
• 
• 

(back to top)

Getting Started

Prerequisites

• Python 3.11
• FastTree
• IQTree
• MAFFT
• HMMER

To install MAFFT:

conda install -c bioconda mafft

Installation

1. Clone the repo

   git clone https://github.com/Bitbol-Lab/Phylogeny-ESM2.git

2. Install the requirements

   pip install -r requirements.txt

(back to top)

Usage

1. You can create a synthetic MSA for a single alignment using:

   python main.py -f <natural_msa_path>

   You can see all available command line arguments with:

   python main.py -h

2. For creating a synthetic MSA for multiple alignments:

   python run.py

   There might be the need to change the extension of the MSA input files used by this script. To do so change run.py:51:

   if f.endswith('.fasta'):

   You can see all available command line arguments with:

   python run.py -h

Results

1. For obtaining hamming distances correlation results:

   python results.py -f <msa_natural_dir> -m <method> -o <msa_synthetic_dir>

   You can see all available command line arguments with:

   python results.py -h

2. For obtaining HMMER scores and their violin plots:

   python hmmer_scores.py -m <method> -s <hmm_profile_dir> -o <msa_dir> -r <output_dir>
   python violin_plot.py <synthetic_scores_dir> <natural_scores_dir>

(back to top)

License

Apache License 2.0

(back to top)