Statistical Modeling for the Analysis of Maldi-tof Mass Spectrometry: Approaches Dealing with Labeled and Label-free Peptide Applications

The analsyses involve novel as well as application oritented statistical methodologies applied to the mass spectrometry based proteomics data. This is a new research area in the statistical point of view, and it requires non-routine statistical modeling approaches. The analyses are based on the assumption of unknown peptide sequences. In such situation, more elaborate statistical modeling approaches need to be developed, to account for the uncertainty caused by the estimation of the masses of the peptides. Two topics are considered in the analyses. The first topic is to separate and quantify overlapping peptides for the label-free applications. The challenge to this topic is to distinguish the isotopic peaks from different peptides which cause an overlap in a spectrum. The second topic aims towards one of the recently developed labeling techniques-- the enzymatic 18O labeling. The key difficulty lies in this topic is how to account for incomplete labeling, non-constant residual variance, etc. Both frequentist and Bayesian approaches are applied. Bayesian analysis becomes the unique choice when non-identifiability issues arise without the incorporation of prior information.