Skip to content

Latest commit

 

History

History
53 lines (44 loc) · 2.17 KB

README.md

File metadata and controls

53 lines (44 loc) · 2.17 KB
Raw

Foldseek_Tweaker 🐰

The Foldseek_Tweaker can adjust the number of alphabet sizes for Foldseek and is helpful for beginners to understand the Foldseek code of training.

Introduction

Foldseek is a breakthrough that transforms the structure of proteins into sequences, greatly accelerating the sequence alignment!

However, it has chosen 20 as its alphabet size for the trade-off between effectiveness and expressiveness. So, imagine we want to pursue one side. Can we adjust this size? Let's try!

Installation

Here are some configurations from our device for reference:

  • CUDA == 11.7
  • Python == 3.9
  • torch == 1.13.0+cu117
  • biopython == 1.83

Please be mindful of version compatibility during your actual setup.

Quick Start

  • Bash Version: sudo bash Encoder.sh tmp/pdb 20
    • 20: Alphabet Size You Want
    • tmp/pdb: Data Directory

The result is named as seqs.csv and is saved in the tmp_20 folder. We have trained 40, 80, 128, and 512 for you to choose from.

Train Your Own Foldseek

  • Bash Version: sudo bash learnAlphabet.sh 20 100 data/pdbs_train.txt data/pdbs_val.txt tmp_20/
    • 20: Alphabet Size You Want
    • 100: Seed for Model Selection
    • data/pdbs_train.txt: Training Data File (See Data Preprocess)
    • data/pdbs_val: Validation Data File (See Data Preprocess)
    • tmp_20/: Output Directory
  • Local: Training Step by Step at one seed
    • Train train_vqvae_local.py
    • Encode encode_pdbs_local.py
    • Evalue create_submat2.py

Data Preprocess

  1. (Optional) Please make sure your protein data has been put in tmp/pdb, or you can run this code on the terminal to test example data. 
    if [ ! -d tmp/pdb ]; then
  curl https://wwwuser.gwdg.de/~compbiol/foldseek/scop40pdb.tar.gz | tar -xz -C tmp
fi
  2. Compile ssw_test on your terminal 
    if [ ! -f tmp/ssw_test ]; then
   git clone --depth 1 https://github.com/mengyao/Complete-Striped-Smith-Waterman-Library tmp/ssw
   (cd tmp/ssw/src && make)
   cp tmp/ssw/src/ssw_test tmp/ssw_test
fi
  3. Create Training Data create_vqvae_training_data_local.py