AdaNovo: Towards Robust \emph{De Novo} Peptide Sequencing in Proteomics against Data Biases (NeurIPS 2024)

Installation

Create a conda environment

First, open the Linux terminal. All of the commands that follow should be entered into this terminal. To create a new conda environment for Adanovo and activate this environment, run the following commands:

conda create --name adanovo python=3.10
conda activate adanovo

Install package

Enter the working directory and install the corresponding environment dependencies for running Adanovo:

cd adanovo_v1
pip install -r requirements.txt

Running

Training

For training a Adanovo model from scratch,

python adanovo.py --mode=train --peak_path=case.mgf --peak_path_val=case.mgf --config=config.yaml --output=log_file/case3

Ensure that your training data (e.g. in. mgf format) is similar in format to the case.mgf provided here. In config.yaml, you can modify the save path of the model, the number of epochs trained, the types of residues in the training data, and the specified AA with PTM types (which must be a subset of the previous residues types)

Validation

For validating a trained model,

python adanovo.py --mode=eval --model=xx.ckpt --peak_path=case.mgf --config=config.yaml --output=log_file/case3

Replace xx.ckpt with the path of the model you have already trained. Replace case.mgf with the dataset you are preparing to validate.

Sequencing

For sequencing peptides from mass spectra in an MGF file

python adanovo.py --mode=denovo --model=xx.ckpt --peak_path=case.mgf --config=config.yaml --output=log_file/case4

Replace xx.ckpt with the path of the model you have already trained. Replace case.mgf with the dataset you are preparing to sequence.

Citation

@inproceedings{
xia2024towards,
title={Towards Robust {\textbackslash}emph\{De Novo\} Peptide Sequencing in Proteomics against Data Biases},
author={Jun Xia and Shaorong Chen and Jingbo Zhou and Xiaojun Shan and Wenjie Du and Zhangyang Gao and Cheng Tan and Bozhen Hu and Jiangbin Zheng and Stan Z. Li },
booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems},
year={2024},
url={https://openreview.net/forum?id=0zfUiSX5si}
}