A signal filtering method for improved quantification and noise discrimination in fourier transform ion cyclotron resonance mass spectrometry-based metabolomics data
- First Online:
- Cite this article as:
- Payne, T.G., Southam, A.D., Arvanitis, T.N. et al. J Am Soc Mass Spectrom (2009) 20: 1087. doi:10.1016/j.jasms.2009.02.001
Direct-infusion electrospray-ionization Fourier transform ion cyclotron resonance mass spectrometry (DI ESI FT-ICR MS) is increasingly being utilized in metabolomics, including the high sensitivity selected ion monitoring (SIM)-stitching approach. Accurate signal quantification and the discrimination of real signals from noise remain major challenges for this approach, with both adversely affected by factors including ion suppression during electrospray, ion—ion interactions in the detector cell, and thermally-induced white noise. This is particularly problematic for complex mixture analysis where hundreds of metabolites are present near the noise level. Here we address relative signal quantification and noise discrimination issues in SIM-stitched DI ESI FT-ICR MS-based metabolomics. Using liver tissue, we first optimized the number of scans (n) acquired per SIM window to address the balance between quantification accuracy versus acquisition time (and thus sample throughput); a minimum of n = 5 is recommended. Secondly, we characterized and computationally-corrected an effect whereby an ion’s intensity is dependent upon its location within a SIM window, exhibiting a 3-fold higher intensity at the high m/z end. This resulted in significantly improved quantification accuracy. Finally, we thoroughly characterized a three-stage filter to discriminate noise from real signals, which comprised a signal-to-noise-ratio (SNR) hard threshold, then a “replicate” filter (retaining only peaks in r-out-of-3 replicate analyses), and then a “sample” filter (retaining only peaks in >s% of biological samples). We document the benefits of three-stage filtering versus one- and two-stage filters, and show the importance of selecting filter parameters that balance the confidence that a signal is real versus the total number of peaks detected.