MSBERT

MSBERT: Embedding Tandem Mass Spectra into Chemically Rational Space by Mask Learning and Contrastive Learning

1. Introduction

In this study, MSBERT based on mask learning and contrastive learning was proposed to get a reasonable embedding representation of MS/MS. MSBERT used the transformer encoder as the backbone and take advantage of the randomness of the mask to construct positive samples for contrastive learning. MSBERT was trained and tested on GNPS dataset. MSBERT had a stronger ability in library matching, with top 1, top5, and top 10 were 0.7871, 0.8950, and 0.9080 on Orbitrap test dataset. The results are significantly better than Spec2Vec and cosine similarity. The rationality of embedding was demonstrated by reducing the dimensionality of the embedding vectors, calculating structural similarity, and spectra clustering.

2. Depends

Anaconda for Python 3.12
Pytorch 2.2

3. Install

1.Install anaconda
2.Install Git
3.Open commond line, create environment and enter with the following commands.

conda create -n MSBERT python=3.12 
conda activate MSBERT

4.Clone the repository and enter.

git clone https://github.com/zhanghailiangcsu/MSBERT.git

5.Install dependency with the following commands.

pip install -r requirements.txt

6.Install Pytorch

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

4. Usage

The MSBERT is public at homepage, every user can download and use it. You can download the trained model on Github. Then refer to the example to use the model for prediction, and directly obtain the embedding vectors from MS/MS. If you have any questions, you can refer to the help documentation or contact the author. Alternatively, you can retrain MSBEERT on your own data refer to example.

import pickle
import torch
import matplotlib.pyplot as plt
from data.ProcessData import MakeTrainData
from model.MSBERTModel import MSBERT
from model.utils import ModelEmbed, ProcessMSP,MSBERTSimilarity

# Load model
model_file = 'model/MSBERT.pkl'
model = MSBERT(100002, 512, 6, 16, 0,100,3)
model.load_state_dict(torch.load(model_file))

# Load demo file
demo_file = 'example/demo.msp'
demo_data,demo_smiles = ProcessMSP(demo_file)
demo_arr = ModelEmbed(model,demo_data,16)

#Calculate the similarity after MSBERT embedding
cos = MSBERTSimilarity(demo_arr,demo_arr)

If you want to use MSBERT web services, you can use the following two methods. The first method is to use the link The second method is to run the following code.

streamlit run Web.py

5.Data available

The filtered GNPS dataset used in this experiment was uploaded to Zenodo.

6. About Spec2Vec

We retrained three Spec2Vec models on the filter GNPS dataset according to instrument type. The training script and file(ipynb) on GNPS Orbitrap set can be accessed at the link. The training data can be accessed at link. The trained model can be downloaded at link. With the uploaded script and training data, it is easy for users to repeat this part of the experiment.

6.Integrate MSBERT into the matchms workflow

We have integrated MSBERT similarity calculation into the matchms workflow. We have uploaded the data and script for the example of integrating MSBERT into the matchms workflow. Users can easily integrate MSBERT into the matchchms process by referring to this section. The example code is as follows.

from matchms_workflow.Similarity import FilterSpectra,MSBERTSimilarity
from matchms.importing import load_from_msp
from matchms import calculate_scores

# Load demo file
demofile = 'matchms_workflow/demo_msms.msp'
spectra = list(load_from_msp(demofile))
spectra = FilterSpectra(spectra)

# Calculate MSBERT similarity scores between all spectra in matchms workflow
MSBERT_Similarity = MSBERTSimilarity()
scores = calculate_scores(references=spectra,
                          queries=spectra,
                          similarity_function=MSBERT_Similarity).to_array()

7.Contact

Hailiang Zhang
E-mail [email protected]