Assessing the inter-instrument variation in clinical metabolic phenotyping workflows

Shaimaa S Khandaker

Metabolomics research involves high-throughput analysis of small molecules found in complex biological fluids, including urine, blood and tissue extracts to study physiological states and chronic disease progression. Metabolomics seeks to capture detailed information of functional metabolic networks that are influenced by genetics, diet, lifestyle and external environment. The main approaches of metabolic phenotyping include targeted and untargeted analysis (also known as metabolic profiling and metabolic fingerprinting). This is achieved with two predominant powerful analytical platforms employed for metabolic phenotyping including nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS) which can detect metabolites to generate high-quality analytical and biological data. Whilst NMR is considered a stable analytical platform that is moderately robust to batch effects, multiple batches of data acquired on liquid chromatography-mass spectrometry (LC-MS) instruments are prone to analytical data variability and thus raising concerns of biological data reproducibility. Efforts in research laboratories to attain equivalent analytical and subsequent biological data across multiple instruments (inter-instrument) on multiple days (inter-day) remain key for the analysis of large-scale cohort studies and the subsequent data processing and interpretation strategies to deal with inter-platform variation. This can improve confidence for the potential discovery of highly crucial and significant biomarkers to identify early stages of various chronic diseases. Therefore, this project initiated an intra-laboratory ring-trial experiment with the aim to evaluate the parallel performance of multiple matched MS instruments. The study evaluated established LC-QTOF-MS methods, run in parallel on three matched instrument set-ups (Waters Acquity UPLC, Bruker Impact II QTOF-MS). The study consisted of two phases: Phase 1a and phase 1b consisted of a multi-instrument comparison of replicate analyses of standards (phase 1a) and known analytes extracted from replicate pooled urine samples (phase 1b); Evaluations for phase 1a and 1b consisted of retention time stability, benchmarking system sensitivity and mass accuracy. Phase 2 consisted of a multi-instrument comparison of a biological sample set consisting of urine collected from 10 males and 10 females. Data evaluation consisted of number of features following peak picking, features <30% RSD in replicate quality control samples, number of features following Bruker pre-processing software, auto MS/MS and comparison of discriminate features generated in multi-variate models. Results from phase 1a and 1b indicated that intra-instrument performance across multiple days was largely consistent., yet variation was observed for inter-instrument comparisons. However, phase 2 indicated that despite inter-instrument variability, the key discriminant features remained consistent across instruments. This study demonstrated good reproducibility of UPLC-QTOF based untargeted metabolomics data and highlighted the importance of repeated analyses using analytical quality control (QC) materials prior to biological sample analysis and contributed to the value of knowledge towards the metabolomics community.

Assessing the inter-instrument variation in clinical metabolic phenotyping workflows

Files and links (1)

Abstract

Details

Metrics