How searching against multiple libraries can lead to biased results in GC/MS-based metabolomics

Area of Science:

• Analytical Chemistry
• Biochemistry
• Computational Biology

Background:

• Electron ionization mass spectra databases are crucial for GC/MS untargeted metabolomics.
• Library search performance is influenced by database characteristics and algorithms.
• The selection of m/z values, often hidden from users, is a critical but overlooked parameter.

Purpose of the Study:

• To investigate how popular software (AMDIS, ChromaTOF, MS Search, Xcalibur) selects m/z values for library searching.
• To evaluate the impact of m/z value selection on untargeted metabolomics library search results.
• To compare the performance of different software packages using real and synthetic mass spectral data.

Main Methods:

• Generated synthetic datasets to analyze m/z value selection in AMDIS, ChromaTOF, MS Search, and Xcalibur.
• Utilized real mass spectral datasets from NIST and FiehnLib libraries.
• Compared library search outcomes across different software using the NIST MS Search Identity algorithm.

Main Results:

• AMDIS and ChromaTOF demonstrated biased library search results under specific conditions.
• Bias occurred when using NIST and FiehnLib libraries concurrently with a scan range below 85, and the target compound was exclusively in FiehnLib.
• MS Search's default m/z selection algorithm avoided these biases.

Conclusions:

• Biased results stem from the absence of scan range information in library metadata.
• AMDIS and ChromaTOF incorrectly treat peaks as missing, penalizing correct identifications.
• MS Search's approach to m/z value selection ensures more reliable untargeted metabolomics analyses.