Skip to content

README.md

Latest commit

 

History

History
41 lines (30 loc) · 1.15 KB

README.md

File metadata and controls

41 lines (30 loc) · 1.15 KB
Raw

CaNAT: Decoding synonymous codon selection with a transformer model

Codon from Amino Acid with a Non-Autoregressive Transformer

Predicts codons and associated confidence score from amino acid sequences. graphical abstract

Installation

git clone https://github.com/Andre-lab/CaNAT.git
cd CaNAT

Create a Conda environment and install necessary packages:

```bash
conda env create -f -n canat

# Activate the environment
conda activate canat
conda install pytorch=2.2.1 -c pytorch
conda install pandas=2.2.1
pip install -e .

Usage / Inference

Run the inference code for an input fasta file and output a prediction file. Example for TEM1 Input file example: TEM1.fasta Output file example: TEM1_prediction/TEM1.csv

# Run the inference script
python network/scripts/CaNAT/inference.py -p network/scripts/CaNAT/parameters_CaNAT.pt -i TEM1.fasta -o TEM1.prediction/

Output is a file with the same number of columns as the input amino acid sequence. The first line lists all codons. Each cell contains the predicted value for a codon at a given position. Confidence scores are calculated by applying the softmax to each row.