Karlijn Hummelink

Chapter 5 - Supplemental material: methods 273 A To ensure that spectral data can be reproducibly generated, the QC/reference samples that were included at the beginning and end of each batch of samples run were used to batch correct the feature values of each batch of samples. Batch correction parameters were determined by comparing the feature values of the QC/reference samples within the batch to gold standard values for the QC/reference sample defined from those contained within the first batch of the development set samples that were run. Note that development set samples and test samples were not used in determining batch correction parameters. Once the batch correction parameters were obtained for a particular batch, all feature values for each spectrum in the batch were corrected. This process adjusted for small m/Z dependent changes in mass spectral performance of the mass spectrometer or between mass spectrometers. The final step in processing of the spectra was another partial ion current normalization step. Using the development set of spectra, a subset of the 274 identified features was defined for final spectral normalization comprising features that had low coefficients of variation and were not associated with the clinical outcome variables in the development set. The m/Z regions (also features) used and fixed as parameters in the fully-specified test are listed in Supplementary Table 6. Supplementary Table 6: m/Z regions used for final normalization Lower limit m/Z Upper limit m/Z 4351.40 4368.27 4372.40 4389.66 4553.30 4577.84 5396.97 5411.16 8353.19 8378.85 8401.71 8420.63 8954.36 8968.33 8968.81 8987.65 8988.02 9009.33 9028.78 9045.84 9089.32 9104.51 9196.31 9219.45 9254.17 9272.44 9639.94 9655.20 10128.04 10151.71 10171.98 10197.16 10197.54 10224.60 14401.51 14460.94

RkJQdWJsaXNoZXIy MTk4NDMw