Authors

Abstract

Multidimensional measurements using state-of-the-art analytical separations and massspectrometry provide great advantages in untargeted analyses for studying bio-chemicalprocesses in a wide range of environmental and biological metabolomics research. However, thelack of rapid analytical methods and robust algorithms to extract knowledge from thismultidimensional and complex data has limited its application. Here, we developed and evaluateda sensitive and high-throughput analytical and computational workflow to enable comprehensiveand accurate metabolite profiling. Our workflow combines liquid chromatography, ion mobilityspectrometry and data-independent acquisition mass spectrometry with PeakDecoder, amachine learning-based algorithm that learns from the raw data and calculates error rates formetabolite identification. We applied PeakDecoder for synthetic biology metabolite profiling ofvarious strains of Pseudomonas putida, Aspergillus pseudoterreus, Aspergillus niger andRhodosporidium toruloides, engineered under projects of the Agile BioFoundry consortium ofnational laboratories. Our results were validated against selected reaction monitoring and gaschromatography platforms, and showed that 2236 metabolite features could be confidently identified and quantified across 116 LC-IM-MS runs from microbial samples using a library builtfrom of 64 standards.