Convex Regression in Multiple Dimensions
- 2010-12-02 (Thu.), 10:30 AM
- Auditorium, 2F, Tsai Yuan-Pei Memorial Hall
- Prof. Peter W. Glynn
- Dept. of Management Science & Engineering, Stanford Univ., USA
Abstract
MALDI?TOF mass spectrometry represents a very fast high?throughput technology for proteomic research. A MALDI?TOF spectrum of a tryptic digest can be acquired in less than one minute and once the sample is spotted on the plate, it can be reanalyzed several times if needed. However, peptide fingerprinting using MALDI?TOF has many limitations. It cannot distinguish between two different sequences of the same mass unless the mass analyzer has sufficient resolution to resolve the two peptides and it cannot identify several proteins in a mixture. We aim to develop better statistical methods to improve MALDI data interpretation with the goal of overcoming some of the limitations of time?of?flight (TOF) analyzers. A mathematical mixture of first hitting time distributions is proposed as a unifying statistical model for the analysis of TOF mass spectrometry data. The model recognizes the time of flight of an ion as a first hitting time and, additionally, models the ion stream as a nonhomogeneous Poisson process. The model guides how a target protein mass spectrum may be deconvolved into signatures of known ions from protein and peptide data bases. The inference methods and ideas are illustrated using a published data set for ovarian cancer.?