AUROMA – Automatic analysis of aroma profiles using machine learning
In dairy research, people are constantly looking for new methods that can help to optimize and control the quality of dairy products. This requires, among other things, a detailed knowledge of all essential aroma components as well as precursors to such.
By: Grith Mortensen
Gas chromatography with mass spectrometric detection (GC-MS) is often used for aroma analyzes. Currently, the data processing of GC-MS data is handled manually, which is very extensive and time consuming especially in large storage attempts. As a result, large trials with GC-MS often focus on already known aroma components, which increases the risk of overlooking information about new biomarkers that might otherwise be available in data. This gives rise to research within optimization of these workflows.
In the AUROMA project, the data analysis from GC-MS is completely automated using machine learning and chemometric data analysis. The automated method extracts far more information from each measurement and at the same time will be more accurate because an automated method does not depend on the user, as is currently the case.
Project period: April 2019 - March 2022
Budget: 2.212.000 DKK
Financing: Milk Levy Fund, Arla Foods, University of Copenhagen
Project manager: Rasmus Bro
Institution: Department of Food Science, University of Copenhagen
Participants: Department of Food Science University of Copenhagen and Arla Foods
Huiwen Yu, Dillen Augustijn, Rasmus Bro. Accelerating PARAFAC2 algorithms for non-negative complex tensor decomposition, Chemometrics and Intelligent Laboratory Systems 214, 15 July 2021, 104312. https://doi.org/10.1016/j.chemolab.2021.104312.