Comprehensive Peak Characterization (CPC) in Untargeted LC–MS Analysis

Abstract

LC–MS-based untargeted metabolomics is heavily dependent on algorithms for automated peak detection and data preprocessing due to the complexity and size of the raw data generated. These algorithms are generally designed to be as inclusive as possible in order to minimize the number of missed peaks. This is known to result in an abundance of false positive peaks that further complicate downstream data processing and analysis. As a consequence, considerable effort is spent identifying features of interest that might represent peak detection artifacts. Here, we present the CPC algorithm, which allows automated characterization of detected peaks with subsequent filtering of low quality peaks using quality criteria familiar to analytical chemists. We provide a thorough description of the methods in addition to applying the algorithms to authentic metabolomics data. In the example presented, the algorithm removed about 35% of the peaks detected by XCMS, a majority of which exhibited a low signal-to-noise ratio. The algorithm is made available as an R-package and can be fully integrated into a standard XCMS workflow.

Comparative metabolomics of root-tips reveals distinct metabolic pathways conferring drought tolerance in contrasting genotypes of rice

“…The mass spectra provided the result data files in CDF formats, which were then imported into the XCMS software using the R software platform ( http://cran.r-project.org ). XCMS could be automatically preprocessed using procedures such peak detection, data baseline filtering, raw sign al extraction, and integration [ 105 ]. Ultimately, after alignment using the statistical component for comparison, the ‘CSV’ file was acquired which included data sets like retention time, sample information, and intensity of peaks.…”

Dual UHPLC-HRMS Metabolomics and Lipidomics and Automated Data Processing Workflow for Comprehensive High-Throughput Gut Phenotyping

“…Prior to further preprocessing, empty spectra were filtered. Next, peak picking was performed with XCMS centWave, and low-quality peaks were filtered using CPC . The CPC step was excluded during extraction and analytical method optimization.…”

“…Next, peak picking was performed with XCMS centWave, 23 and low-quality peaks were filtered using CPC. 17 The CPC step was excluded during extraction and analytical method optimization. The remaining peaks were merged with neighboring peaks within each sample, and RT alignment was performed using XCMS OBI-Warp.…”

“…16 The second challenge in untargeted preprocessing is the retrieval of false positive peaks, which add noise and might result in falsely retained discriminant features in downstream comparisons. 13 To circumvent this, the comprehensive peak characterization (CPC) 17 algorithm characterizes and filters peaks after (1). Parameters for XCMS were optimized using AutoTuner, IPO, and analyst knowledge (2).…”

Picky with peakpicking: assessing chromatographic peak quality with simple metrics in metabolomics

“…Noisy data and imperfect detection algorithms introduce a tradeoff between false positives (where contamination, background instrument or chemical noise is misclassified as biological signal) and false negatives (where real signals are undetected). Existing algorithms tend to favor the inclusion of false positives because downstream analyses can always remove erroneous mass features, but false negatives cannot be later recovered [ 4 , 5 ]. However, this approach requires more time from the researcher as they manually evaluate a potentially enormous number of mass features (MFs), a task that scales combinatorially with the number of samples and compounds measured [ 6 ].…”