Impact of Preprocessing and data analysis strategies on metabolite annotation in biological samples: A mass spectrometry-based metabolomics study using feature-based molecular networking.

Metabolomics workflows involve multiple complex steps including sample collection, storage, preparation, metabolite extraction, analytical platforms selection, data acquisition and interpretation. Each step may introduce variability that affects the quality and reliability of metabolomic data. To systematically investigate the effects of these factors on metabolomics outcomes, plasma samples from four different anatomical sites of colon cancer patients were analyzed using Liquid chromatography- Quadrupole-Exactive Orbitrap mass spectrometry (LC-Q-Exactive Orbitrap MS) for untargeted metabolomics. Response surface methodology was employed to optimize the ultrasound-assisted extraction conditions during sample pretreatment. Data analysis strategies were systematically evaluated, including Feature-Based Molecular Networking (FBMN) construction parameters and comparative assessment of different FBMN platforms for metabolite annotation. The optimized extraction conditions were determined as 300 % methanol concentration, sample freezing at -20 °C for 40 min, followed by ultrasonication for 5 min. Sample standardization protocols requiring single-use portioning and limiting freeze-thaw cycles to ≤2-3 cycles were identified as essential for reliable biomarker discovery and therapeutic mechanism exploration. Optimal FBMN construction parameters comprised a 25-min gradient elution time, 50 mm chromatographic column length, and high sample concentration. Comparative evaluation of Global Natural Products Social Molecular Networking (GNPS) and MZmine implementations of FBMN revealed that GNPS was recommended for studies prioritizing comprehensive annotation coverage and discovery-oriented metabolomics, while MZmine was preferred for method development, or applications requiring local processing without external data upload. This study demonstrated that preprocessing and data analysis strategies were critical determinants of data quality in untargeted plasma metabolomics. The findings provided evidence-based recommendations for experimental design, storage conditions, and data handling procedures that can guide protocol standardization and minimize undesired analytical variation in metabolomics studies.