An Automated Peak Identification/Calibration Procedure forHigh-Dimensional Protein Measures From Mass Spectrometers

摘要

Discovery of “signature” protein profiles that distinguishdisease states (eg, malignant, benign, and normal) is a key steptowards translating recent advancements in proteomic technologiesinto clinical utilities. Protein data generated from massspectrometers are, however, large in size and have complexfeatures due to complexities in both biological specimens andinterfering biochemical/physical processes of the measurementprocedure. Making sense out of such high-dimensional complexdata is challenging and necessitates the use of a systematic dataanalytic strategy. We propose here a data processing strategy fortwo major issues in the analysis of suchmass-spectrometry-generated proteomic data: (1) separation ofprotein “signals” from background “noise” in proteinintensity measurements and (2) calibration of protein mass/chargemeasurements across samples. We illustrate the two issues andthe utility of the proposed strategy using data from a prostatecancer biomarker discovery project as an example.