(free for academic use)

3D Molecular Network for Mass Spectra Prediction (3DMolMS) is a deep neural network model to predict the MS/MS spectra of compounds from their 3D conformations. This model's molecular representation, learned through MS/MS prediction tasks, can be further applied to enhance performance in other molecular-related tasks, such as predicting retention times and collision cross sections.

Read our paper in Bioinformatics | Try our online service at GNPS | Install from PyPI

3DMolMS is available on PyPI. You can install the latest version using pip:

pip install molnetpack

# PyTorch must be installed separately. 
# For CUDA 11.6, install PyTorch with the following command:
pip install torch==1.13.0+cu116 torchvision==0.14.0+cu116 torchaudio==0.13.0 --extra-index-url https://download.pytorch.org/whl/cu116

# For CUDA 11.7, use:
pip install torch==1.13.0+cu117 torchvision==0.14.0+cu117 torchaudio==0.13.0 --extra-index-url https://download.pytorch.org/whl/cu117

# For CPU-only usage, use:
pip install torch==1.13.0+cpu torchvision==0.14.0+cpu torchaudio==0.13.0 --extra-index-url https://download.pytorch.org/whl/cpu

3DMolMS can also be installed through source codes:

git clone https://github.com/JosieHong/3DMolMS.git
cd 3DMolMS

pip install .

To get started quickly, you can load a CSV or MGF file to predict MS/MS and then plot the predicted results.

import torch
from molnetpack import MolNet

# Set the device to CPU for CPU-only usage:
device = torch.device("cpu")

# For GPU usage, set the device as follows (replace '0' with your desired GPU index):
# gpu_index = 0
# device = torch.device(f"cuda:{gpu_index}")

# Instantiate a MolNet object
molnet_engine = MolNet(device, seed=42) # The random seed can be any integer. 

# Load input data (here we use a CSV file as an example)
molnet_engine.load_data(path_to_test_data='./test/input_msms.csv')
# molnet_engine.load_data(path_to_test_data='./test/input_msms.mgf') # MGF file is also supported

# Predict MS/MS
spectra = molnet_engine.pred_msms(path_to_results='./test/output_msms.mgf')

# Plot the predicted MS/MS with 3D molecular conformation
molnet_engine.plot_msms(dir_to_img='./img/')

For CCS prediction, please use the following codes after instantiating a MolNet object.

# Load input data
molnet_engine.load_data(path_to_test_data='./test/input_ccs.csv')

# Pred CCS
ccs_df = molnet_engine.pred_ccs(path_to_results='./test/output_ccs.csv')

For saving the molecular embeddings, please use the following codes after instantiating a MolNet object.

# Load input data
molnet_engine.load_data(path_to_test_data='./test/input_savefeat.csv')

# Inference to get the features
ids, features = molnet_engine.save_features()

print('Titles:', ids)
print('Features shape:', features.shape)

The sample input files, a CSV and an MGF, are located at ./test/demo_input.csv and ./test/demo_input.mgf, respectively. If the input data is only expected to be used in CCS prediction, you may assign an arbitrary numerical value to the Collision_Energy field in the CSV file or to COLLISION_ENERGY in the MGF file. It's important to note that during the data loading phase, any input formats that are not supported will be automatically excluded. Below is a table outlining the types of input data that are supported:

The sample output files are at ./test/demo_msms.mgf and ./test/demo_ccs.csv. Below is an example of a predicted MS/MS spectrum plot.

The documents for running MS/MS prediction from source codes are at MSMS_PRED.md.

Step 0: Clone the Repository and Set Up the Environment

Clone the 3DMolMS repository and install the required packages using the following commands:

git clone https://github.com/JosieHong/3DMolMS.git
cd 3DMolMS

# Please install the packages if you have not installed them yet. 
pip install .

Step 1: Obtain the Pretrained Model

Download the pretrained model (molnet_pre_etkdgv3.pt.zip) from Google Drive or train the model yourself. For details on pretraining the model on the QM9 dataset, refer to PRETRAIN.md.

Step 2: Prepare the Datasets

Download and organize the datasets into the ./data/ directory. The current version uses four datasets:

1. Agilent DPCL, provided by Agilent Technologies.
2. NIST20, available under license for academic use.
3. MoNA, publicly available.
4. Waters QTOF, our own experimental dataset.

The data directory structure should look like this:

|- data
  |- origin
    |- Agilent_Combined.sdf
    |- Agilent_Metlin.sdf
    |- hr_msms_nist.SDF
    |- MoNA-export-All_LC-MS-MS_QTOF.sdf
    |- MoNA-export-All_LC-MS-MS_Orbitrap.sdf
    |- waters_qtof.mgf

Step 3: Preprocess the Datasets

Run the following commands to preprocess the datasets. Specify the dataset with --dataset and select the instrument type as qtof. Use --maxmin_pick to apply the MaxMin algorithm for selecting training molecules; otherwise, selection will be random. The dataset configurations are in ./src/molnetpack/config/preprocess_etkdgv3.yml.

python ./src/scripts/preprocess.py --dataset agilent nist mona waters \
--instrument_type qtof \
--data_config_path ./src/molnetpack/config/preprocess_etkdgv3.yml \
--mgf_dir ./data/mgf_debug/

Step 4: Train the Model

Use the following commands to train the model. Configuration settings for the model and training process are located in ./src/molnetpack/config/molnet.yml.

python ./src/scripts/train.py --train_data ./data/qtof_etkdgv3_train.pkl \
--test_data ./data/qtof_etkdgv3_test.pkl \
--model_config_path ./src/molnetpack/config/molnet.yml \
--data_config_path ./src/molnetpack/config/preprocess_etkdgv3.yml \
--checkpoint_path ./check_point/molnet_qtof_etkdgv3.pt \
--transfer --resume_path ./check_point/molnet_pre_etkdgv3.pt

Additional application

3DMolMS is also capable of predicting molecular properties and generating reference libraries for molecular identification. Examples of such applications include retention time prediction and collision cross-section prediction. For more details, refer to PROP_PRED.md and GEN_REFER_LIB.md respectively.

If you use 3DMolMS in your research, please cite our paper:

@article{hong20233dmolms,
  title={3DMolMS: prediction of tandem mass spectra from 3D molecular conformations},
  author={Hong, Yuhui and Li, Sujun and Welch, Christopher J and Tichy, Shane and Ye, Yuzhen and Tang, Haixu},
  journal={Bioinformatics},
  volume={39},
  number={6},
  pages={btad354},
  year={2023},
  publisher={Oxford University Press}
}

Thank you for considering 3DMolMS for your research needs!

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.