Du Peicheng, Stolovitzky Gustavo, Horvatovich Peter, Bischoff Rainer, Lim Jihyeon, Suits Frank
IBM Computational Biology Center, P.O. Box 218, Yorktown Heights, NY 10598, USA.
Bioinformatics. 2008 Apr 15;24(8):1070-7. doi: 10.1093/bioinformatics/btn078. Epub 2008 Mar 18.
Mass spectrometry data are subjected to considerable noise. Good noise models are required for proper detection and quantification of peptides. We have characterized noise in both quadrupole time-of-flight (Q-TOF) and ion trap data, and have constructed models for the noise.
We find that the noise in Q-TOF data from Applied Biosystems QSTAR fits well to a combination of multinomial and Poisson model with detector dead-time correction. In comparison, ion trap noise from Agilent MSD-Trap-SL is larger than the Q-TOF noise and is proportional to Poisson noise. We then demonstrate that the noise model can be used to improve deisotoping for peptide detection, by estimating appropriate cutoffs of the goodness of fit parameter at prescribed error rates. The noise models also have implications in noise reduction, retention time alignment and significance testing for biomarker discovery.
质谱数据存在大量噪声。为了准确检测和定量肽段，需要良好的噪声模型。我们已对四极杆飞行时间（Q-TOF）和离子阱数据中的噪声进行了特征描述，并构建了噪声模型。
我们发现，应用生物系统公司QSTAR的Q-TOF数据中的噪声与带有探测器死时间校正的多项分布和泊松模型的组合拟合良好。相比之下，安捷伦MSD-Trap-SL的离子阱噪声大于Q-TOF噪声，且与泊松噪声成正比。然后，我们证明了通过在规定错误率下估计拟合优度参数的适当截止值，噪声模型可用于改进肽段检测的去同位素化。噪声模型在降噪、保留时间校准以及生物标志物发现的显著性检验中也具有重要意义。
