Karlijn Hummelink

Chapter 5 - Supplemental material: methods 267 A adjustment of alignment in the m/Z axis, background was subtracted from the raster spectra and peaks are located. This peak list was then used to align raster spectra prior to background subtraction, using a set of 41 fixed alignment points, Supplementary Table 1. Supplementary Table 1: Points in m/Z used to align the raster spectra 3168 7202 8919 12173 15127 28298 4153 7563 8994 12572 15263 4183 7614 9133 12864 15869 4792 7934 9310 13555 17253 5773 8034 9427 13763 18630 5802 8206 10739 13882 21066 6433 8684 10938 14040 23024 6631 8812 11527 14405 28090 The raster spectra were then filtered to keep only spectra where at least 20 peaks were found and at least 5 of the alignment points in Supplementary Table 2 were used in their alignment. Deep MALDI, 400,000 laser-shot averages were then created by averaging together 500 randomly selected from the pool of filtered raster spectra. b. Processing of Deep MALDI Average Spectra “Spectrum” refers to a Deep MALDI average spectrum. The background in each spectrum was estimated and subtracted. Spectra were then normalized in several stages. First, a coarse normalization was performed using a partial ion current (PIC) approach using regions of the spectra that showed low variability across the population of interest and that showed no sign of association with outcomes. Spectral regions known to be intrinsically unstable were excluded from the regions used for normalization. The regions used in this normalization step were defined using the development set of spectra and then held fixed as fully specified parameters of the test. For each spectrum the area under the spectra within each m/Z region used for normalization was calculated and summed to produce a normalization coefficient. Each spectrum was then divided by its own normalization coefficient to yield the normalized spectrum. Supplementary Table 2 gives the upper and lower limits in m/Z of the regions used in normalization.

RkJQdWJsaXNoZXIy MTk4NDMw