Master Thesis Code: Prediction and Visualization of Molecular Substructures using Mass Spectrometry

This repository contains the code, benchmarks, visualizations, and experiments for my master’s thesis in Media Informatics at Hochschule Düsseldorf:

Prediction and Visualization of Molecular Substructures using Mass Spectrometry
Kevin R. Zielke, 2025
Supervised by Prof. Dr. Florian Huber, Advisors David Sokalski, M.Sc.
DOI: 10.13140/RG.2.2.20175.34725

The work focuses on improving the interpretation of molecular similarity in mass spectrometry by introducing Stacked Similarity Maps (SSM) and related metrics for group-based similarity assessment in chemical space exploration.

This repository primarily serves as:

• A transparent record of the experimental methodology
• A reference implementation for the concepts introduced in the thesis
• Supplementary material for readers of the thesis

---

Reproducibility & Usage

The notebooks have some dependencies to code I implemented in the backend of the ms_chemical_space_explorer web application (which contains a cleaner and more modular implementation than this exploratory research code). Unfortunately this project is currently under development and private. Therefore you can't run the notebooks yourself. However this might change in the future, so here is how you would go about it:

Setup

Download compounds_ms2structures.csv and biostructures_combined.csv from Zenodo and place them under data/datasets/.

To create and activate the conda environment:

conda env create -f chemspace.yml
conda activate chemspace

This environment includes the required Python version, RDKit, and the scientific Python stack used in the experiments. The notebooks create some files on disk and are intended to run sequentially from 00_biostructures_dataset_selection.ipynb to 06_example_vis.ipynb.