Tandem Mass Spectrum Prediction with Graph Transformers

Overview

MassFormer

This is the original implementation of MassFormer, a graph transformer for small molecule MS/MS prediction. Check out the preprint on arxiv.

Setting Up Environment

We recommend using conda. Three conda yml files are provided in the env/ directory (cpu.yml, cu101.yml, cu102.yml), providing different pytorch installation options (CPU-only, CUDA 10.1, CUDA 10.2). They can be trivially modified to support other versions of CUDA.

To set up an environment, run the command conda env create -f ${CONDA_YAML}, where ${CONDA_YAML} is the path to the desired yaml file.

Downloading NIST Data

Note: this step requires a Windows System or Virtual Machine

The NIST 2020 LC-MS/MS dataset can be purchased from an authorized distributor. The spectra and associated compounds can be exported to MSP/MOL format using the included lib2nist software. There is a single MSP file which contains all of the mass spectra, and multiple MOL files which include the molecular structure information for each spectrum (linked by ID). We've included a screenshot describing the lib2nist export settings.

Alt text

There is a minor bug in the export software that sometimes results in errors when parsing the MOL files. To fix this bug, run the script python mol_fix.py ${MOL_DIR}, where ${MOL_DIR} is a path to the NIST export directory with MOL files.

Downloading Massbank Data

The MassBank of North America (MB-NA) data is in MSP format, with the chemical information provided in the form of a SMILES string (as opposed to a MOL file). It can be downloaded from the MassBank website, under the tab "LS-MS/MS Spectra".

Exporting and Preparing Data

We recommend creating a directory called data/ and placing the downloaded and uncompressed data into a folder data/raw/.

To parse both of the datasets, run parse_and_export.py. Then, to prepare the data for model training, run prepare_data.py. By default the processed data will end up in data/proc/.

Setting Up Weights and Biases

Our implementation uses Weights and Biases (W&B) for logging and visualization. For full functionality, you must set up a free W&B account.

Training Models

A default config file is provided in "config/template.yml". This trains a MassFormer model on the NIST HCD spectra. Our experiments used systems with 32GB RAM, 1 Nvidia RTX 2080 (11GB VRAM), and 6 CPU cores.

The config/ directory has a template config file template.yml and 8 files corresponding to the experiments from the paper. The template config can be modified to train models of your choosing.

To train a template model without W&B with only CPU, run python runner.py -w False -d -1

To train a template model with W&B on CUDA device 0, run python runner.py -w True -d 0

Reproducing Tables

To reproduce a model from one of the experiments in Table 2 or Table 3 from the paper, run python runner.py -w True -d 0 -c ${CONFIG_YAML} -n 5 -i ${RUN_ID}, where ${CONFIG_YAML} refers to a specific yaml file in the config/ directory and ${RUN_ID} refers to an arbitrary but unique integer ID.

Reproducing Visualizations

The explain.py script can be used to reproduce the visualizations in the paper, but requires a trained model saved on W&B (i.e. by running a script from the previous section).

To reproduce a visualization from Figures 2,3,4,5, run python explain.py ${WANDB_RUN_ID} --wandb_mode=online, where ${WANDB_RUN_ID} is the unique W&B run id of the desired model's completed training script. The figues will be uploaded as PNG files to W&B.

Reproducing Sweeps

The W&B sweep config files that were used to select model hyperparameters can be found in the sweeps/ directory. They can be initialized using wandb sweep ${PATH_TO_SWEEP}.

You might also like...
Multivariate Time Series Forecasting with efficient Transformers. Code for the paper
Multivariate Time Series Forecasting with efficient Transformers. Code for the paper "Long-Range Transformers for Dynamic Spatiotemporal Forecasting."

Spacetimeformer Multivariate Forecasting This repository contains the code for the paper, "Long-Range Transformers for Dynamic Spatiotemporal Forecast

[CIKM 2019] Code and dataset for "Fi-GNN: Modeling Feature Interactions via Graph Neural Networks for CTR Prediction"

FiGNN for CTR prediction The code and data for our paper in CIKM2019: Fi-GNN: Modeling Feature Interactions via Graph Neural Networks for CTR Predicti

Few-Shot Graph Learning for Molecular Property Prediction

Few-shot Graph Learning for Molecular Property Prediction Introduction This is the source code and dataset for the following paper: Few-shot Graph Lea

This is the repo for the paper `SumGNN: Multi-typed Drug Interaction Prediction via Efficient Knowledge Graph Summarization'. (published in Bioinformatics'21)

SumGNN: Multi-typed Drug Interaction Prediction via Efficient Knowledge Graph Summarization This is the code for our paper ``SumGNN: Multi-typed Drug

Unofficial TensorFlow implementation of Protein Interface Prediction using Graph Convolutional Networks.
Unofficial TensorFlow implementation of Protein Interface Prediction using Graph Convolutional Networks.

[TensorFlow] Protein Interface Prediction using Graph Convolutional Networks Unofficial TensorFlow implementation of Protein Interface Prediction usin

Cancer Drug Response Prediction via a Hybrid Graph Convolutional Network
Cancer Drug Response Prediction via a Hybrid Graph Convolutional Network

DeepCDR Cancer Drug Response Prediction via a Hybrid Graph Convolutional Network This work has been accepted to ECCB2020 and was also published in the

MVGCN: a novel multi-view graph convolutional network (MVGCN) framework for link prediction in biomedical bipartite networks.

MVGCN MVGCN: a novel multi-view graph convolutional network (MVGCN) framework for link prediction in biomedical bipartite networks. Developer: Fu Hait

Code for the paper Relation Prediction as an Auxiliary Training Objective for Improving Multi-Relational Graph Representations (AKBC 2021).
Code for the paper Relation Prediction as an Auxiliary Training Objective for Improving Multi-Relational Graph Representations (AKBC 2021).

Relation Prediction as an Auxiliary Training Objective for Knowledge Base Completion This repo provides the code for the paper Relation Prediction as

Official implementation of "Motif-based Graph Self-Supervised Learning forMolecular Property Prediction"

Motif-based Graph Self-Supervised Learning for Molecular Property Prediction Official Pytorch implementation of NeurIPS'21 paper "Motif-based Graph Se

Comments
  • How to label the fragment type of NIST20?

    How to label the fragment type of NIST20?

    Hi,

    Thanks for the great work! When I tried to implement it on my server, I met some problems with splitting the data of 'HCD' and 'CID'.

    Following the instructor in README, I exported lr_msms_nist and hr_msms_nist of NIST20. However, after parse_and_export.py, the fragment type (frag_mode) of them is NaN. So how do you label the fragment type of them?

    Thanks, Josie

    opened by JosieHong 0
Owner
Röst Lab
Röst lab at U of T -- join us at https://gitter.im/Roestlab/Lobby
Röst Lab
Implementation for the IJCAI2021 work "Beyond the Spectrum: Detecting Deepfakes via Re-synthesis"

Beyond the Spectrum Implementation for the IJCAI2021 work "Beyond the Spectrum: Detecting Deepfakes via Re-synthesis" by Yang He, Ning Yu, Margret Keu

Yang He 27 Jan 7, 2023
Use MATLAB to simulate the signal and extract features. Use PyTorch to build and train deep network to do spectrum sensing.

Deep-Learning-based-Spectrum-Sensing Use MATLAB to simulate the signal and extract features. Use PyTorch to build and train deep network to do spectru

null 10 Dec 14, 2022
This code is a near-infrared spectrum modeling method based on PCA and pls

Nirs-Pls-Corn This code is a near-infrared spectrum modeling method based on PCA and pls 近红外光谱分析技术属于交叉领域,需要化学、计算机科学、生物科学等多领域的合作。为此,在(北邮邮电大学杨辉华老师团队)指导下

Fu Pengyou 6 Dec 17, 2022
Frequency Spectrum Augmentation Consistency for Domain Adaptive Object Detection

Frequency Spectrum Augmentation Consistency for Domain Adaptive Object Detection Main requirements torch >= 1.0 torchvision >= 0.2.0 Python 3 Environm

null 15 Apr 4, 2022
Aiming at the common training datsets split, spectrum preprocessing, wavelength select and calibration models algorithm involved in the spectral analysis process

Aiming at the common training datsets split, spectrum preprocessing, wavelength select and calibration models algorithm involved in the spectral analysis process, a complete algorithm library is established, which is named opensa (openspectrum analysis).

Fu Pengyou 50 Jan 7, 2023
Price-Prediction-For-a-Dream-Home - A machine learning based linear regression trained model for house price prediction.

Price-Prediction-For-a-Dream-Home ROADMAP TO THIS LINEAR REGRESSION BASED HOUSE PRICE PREDICTION PREDICTION MODEL Import all the dependencies of the p

DIKSHA DESWAL 1 Dec 29, 2021
Doge-Prediction - Coding Club prediction ig

Doge-Prediction Coding Club prediction ig Basically: Create an application that

null 1 Jan 10, 2022
Dense Prediction Transformers

Vision Transformers for Dense Prediction This repository contains code and models for our paper: Vision Transformers for Dense Prediction René Ranftl,

Intel ISL (Intel Intelligent Systems Lab) 1.3k Dec 28, 2022
Code for Transformers Solve Limited Receptive Field for Monocular Depth Prediction

Official PyTorch code for Transformers Solve Limited Receptive Field for Monocular Depth Prediction. Guanglei Yang, Hao Tang, Mingli Ding, Nicu Sebe,

stanley 152 Dec 16, 2022
Dense Prediction Transformers

Vision Transformers for Dense Prediction This repository contains code and models for our paper: Vision Transformers for Dense Prediction René Ranftl,

Intelligent Systems Lab Org 1.3k Jan 2, 2023