chlange/RamanSpectraBioprocessAnalystesTec5

Dataset Overview

This dataset contains Raman spectra of mixtures of glucose, sodium acetate, and magnesium sulfate. It is part of a series of 8 datasets that use eight different spectrometers that measure nearly the same samples. Some datasets have a bit more samples than others. Each spectrum is paired with ground truth concentration labels verified by enzymatic assays, reflecting the concentration ranges typically found in E. coli fermentation processes.

Target Parameters and Concentration Ranges

The dataset contains measured Raman spectra of samples with different parameters from the following substances:

• —Glucose
• —Acetate
• —Magnesium Sulfate

Reference values for the samples were measured using an HT analyzer (Cedex BioHT, Roche Diagnostics GmbH, Mannheim, Germany).

Data Acquisition

Raman spectra were recorded using the follwing settings:

• —Instrument: Tec5 Multi-Spec© Raman
• —Laser Wavelength: 785 nm
• —Exposure Time: 12 s
• —Laser Power: 500 mW
• —Scans per Sample: 1
• —Number of Samples: 76
• —Container Material: Plastic

Citation

Users should cite the original publication when using this dataset (Lange et. al. https://doi.org/10.1016/j.saa.2025.125861)

or BibTex:

@article{
  LANGE2025125861,
  title = {Comparing machine learning methods on Raman spectra from eight different spectrometers},
  journal = {Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy},
  volume = {334},
  pages = {125861},
  year = {2025},
  issn = {1386-1425},
  doi = {https://doi.org/10.1016/j.saa.2025.125861},
  url = {https://www.sciencedirect.com/science/article/pii/S1386142525001672},
  author = {Christoph Lange and Maxim Borisyak and Martin Kögler and Stefan Born and Andreas Ziehe and Peter Neubauer and M. Nicolas Cruz Bournazou},
  keywords = {Raman spectroscopy, Machine learning, Partial least squares, Convolutional neural network},
}