@inproceedings{01HA6SDDJ67RVCKFPC6ZB65MH0,
  abstract     = {{In light of climate change and shifting consumer perception, biological production processes such as microbial bioconversion and fermentation are gaining more traction for the production of drop-in chemicals. However, despite efforts in strain engineering and bioprocess development, biological production often remains at a cost disadvantage compared to its well-established chemical counterpart. Although the advent of a multitude of omics technologies allowed a deeper understanding of the microbial metabolism, these insights are only validated at lab scale. The fermentation process is too often still inadequately understood. Indeed, initial cultivation conditions and changing environmental dynamics during fermentation can lead to an unpredicted arrest of growth or production if not monitored well and associated, a high market cost of these biochemicals.

Traditional workflows are too time- and resource consuming and analysis-limited to explain out-of-specification changing dynamics in metabolism during the process. Therefore, in this work, we employed laser-assisted rapid evaporative ionization mass spectrometry (LA-REIMS) to profile a biosurfactant-producing microbial culture in real-time in a matter of seconds. Although beneficial from a rapid sampling perspective, the lack of retention time, peak shape and sub-ppm mass accuracy in ambient ionization mass spectrometry complicates feature identification. For example, considerable mass shifts caused by instrument instability during prolonged analyses in longitudinal monitoring of multi-day processes and between different batch analyses complicates interpretation of the results. Therefore, we implemented an improved workflow compared to the proprietary processing software for REIMS data. To mitigate the inter- and intravariability during prolonged analyses, we developed a processing and alignment workflow which does not necessitate known lock mass compounds in an untargeted setting. The workflow is expanded to include dynamic subtractive noise removal and allows for putative feature and adduct identification of potentially relevant biomarkers indicative of a performant biological process. It will allow to develop machine learning models to effectively predict a diverse set of changes in the performance of the fermentation process under different cultivation conditions, and thus, in the end, to mitigate these changes in performance at-line and to develop more cost-efficient bioprocesses.}},
  author       = {{de Fooz, Nicolas and De Graeve, Marilyn and Vangeenderhuysen, Pablo and Maes, Karolien and Roelants, Sophie and De Maeseneire, Sofie and Soetaert, Wim and Vanhaecke, Lynn}},
  booktitle    = {{YoungNMC Symposium 2023, Abstracts}},
  keywords     = {{metabolomics,untargeted metabolomics,biosurfactants,sophorolipids,fermentation,machine learning}},
  language     = {{eng}},
  location     = {{Nijmegen, Netherlands}},
  title        = {{Advancing ambient ionisation mass spectrometry towards at-line microbial fermentation monitoring}},
  year         = {{2023}},
}