Liquid chromatography-mass spectrometry (LC-MS) has revolutionised metabolic phenotyping studies, yet spectral data acquired from these instruments pose significant challenges due to their susceptibility to sample and instrument variations. Consequently, ensuring data quality and reliability has become a time-consuming bottleneck in major data processing pipelines, requiring manual in-person assessment of each sample. To address this, we present an analytical methodology employing total ion current (TIC) for rapid sample quality and instrument performance evaluation.
Our study utilised an exploratory dataset comprising of 3439 control samples obtained using LC-Quadrupole Time-of-Flight (QToF)-MS instruments. These samples include quality control (matrix-free) and calibration samples of different concentration levels along with long-term biological references of different matrices. The mass-to-charge (m/z) values stored within the spectrum array were filtered against 73 known compounds of interest that encompass calibration references, internal standards, and endogenous analytes.
Our analytical workflow involved a preliminary check on the spectral data available, followed by univariate and multivariate statistical analyses, including violin plots, time series, and principal component analyses. These models assessed TIC values across chromatographic regions, revealing sample quality and instrument variations through the identification of significant outliers using deviation-based threshold limits. Further examination of the TIC of extracted ions enabled observation of compound stability and detection of mass accuracy drift.
Our study addresses a significant issue in metabolic phenotyping by introducing a novel LC-MS data quality assessment methodology. By integrating TIC values with statistical modelling approaches, we offer a swift systematic approach to evaluate sample quality and instrument performance without the need for manual assessment. Through automation and reporting features, the methodology significantly lowers the time and resources required while ensuring data quality and reliability. By mitigating the impact of sample and instrument variances, we can facilitate more accurate and reproducible metabolic phenotyping studies, paving the way for deeper insights into biological systems.