Mass Spectra

Article

https://aile3.ijs.si/dunja/SiKDD2023/Papers/IS_2023_-_SIKDD_paper_11.pdf

Purpose

Create a pipeline that extracts various molecule fingerprints and their embeddings from mass spectra data. Then train ML models to predict the fingerprint from the embedding.

Usage

1. Install

• Install Maven (needed for scyjava python package) (install guide) and add it to PATH
• Install conda (install guide)
• Run conda env create -f conda.yml --name mass_spectra to create conda environment called mass_spectra (note there might be some non fatal errors for some packages)
• Run conda activate mass_spectra to activate the environment
• Run pip install -r requirements.txt to install python packages that might be skipped by conda

2. Setup

• Extract files from data to source/dataset/ folder. (NOTE: the source/dataset/ should directly contain the extracted files without any subfolders).
• Follow the jupyter notebooks in pipeline/ folder. (NOTE: the notebooks should be run in the order they are numbered).
• If you follow the jupyter notebooks the basic pipeline will be executed and all embeddings, Spec2Vec models and ML model should be generated along with evaluation files.

Project Structure

• helper/ - contains helper functions used by the project
• mass_spectra - main command line tools also used in jupyter notebooks
• pipeline/ - main pipeline built in jupyter notebooks from embedding to evaluation of trained models
• playground/ - contains jupyter notebooks used for testing and playing around
• source/ - contains the dataset and the generated files (models, spectra, embeddings, etc.)
• conda.yml - contains the conda environment setup
• requirements.txt - contains the python packages that are not installed by conda as a fallback

General Information

1. Articles

• https://jcheminf.biomedcentral.com/articles/10.1186/s13321-022-00636-1
• https://www.sciencedirect.com/science/article/pii/S2352340923002573
• https://jcheminf.biomedcentral.com/articles/10.1186/s13321-017-0220-4
• https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1008724

2. Other important external links

• https://cdk.github.io/
• http://cdk.github.io/cdk/latest/docs/api/index.html
• https://github.com/iomega/spec2vec

3. Terminology

• InChikey = International Chemical Identifier key
• TBDMS = Tert-Butyldimethylsilyl chloride
• TMS = Tetramethylsilane
• RAW = raw data, BS = preprocessed data
• GC–MS-EI = Gas Chromatography–Mass Spectrometry (GC–MS) with Cold Electron Ionization (EI)
• Molecular fingerprints = bit string representation
• SMILES = Simplified molecular-input line-entry system
• metabolomics = the large-scale study of small molecules, commonly known as metabolites, within cells, biofluids, tissues or organisms
• metabolites = a substance made or used when the body breaks down food, drugs or chemicals, or its own tissue
• spec2vec = embedding approach that utalizes word2vec to create embeddings from spectral data
• precursor (ion) = ion which is the source of a fragmentation either spontaneous or induced by collisions. Also known as "mother ion".
• m/z = M stands for mass and Z stands for charge number of ions. m/z is the mass-to-charge ratio. (Z is often 1, so m/z is often the same as mass.)
• Tanimoto similartiy = measure of similarity between two sets of data. It is a metric used to compare the similarity of two sets of data, and is often used in machine learning and data science.

4. Chemistry Development Kit (CDK)

• built in Java
• python wrapper: scyjava
• java and python wrapper need Maven

from scyjava import config, jimport
config.endpoints.append('org.openscience.cdk:cdk-bundle:2.8')

CircularFingerprinter = jimport('org.openscience.cdk.fingerprint.CircularFingerprinter')