Seminar in applied mathematics and statistics

SPEAKER: Ron Wehrens, Biometris / Biosciences, Wageningen UR

TITLE: Fast alignment of peak lists with Parametric Time Warping

ABSTRACT:
Measurements on biological and medical samples are nowadays often done using so-called hyphenated techniques, where the chemical components of the sample are first separated using a form of chromatography, and then measured in a detection step. Identification then is done by looking at spectral characteristics as well as retention time (the time needed to reach the detector in the first place). Unfortunately, these retention times can be variable, and there is a real danger that features are mislabeled.
Obviously, in such cases statistical analysis of the data is doomed to fail.
A large number of methods have been proposed to correct these time shifts. Parametric Time Warping (PTW) is one of them. It has extremely good properties to tackle specific types of shifts, often occuring in practice, but like most other alignment algorithms can be quite slow. In this talk I will discuss a new development in PTW, where instead of the usual chromatographic profiles, the shift correction is calculated from the peaks. This not only reduces computation times (and memory consumption) significantly, but also opens up whole new applications in the areas of mass-spectrometry-based metabolomics. The improved speed also allows to replace the usual gradient-based optimization methods with others that are less likely to be trapped in a local optimum, leading to more consistent alignment results.

Tea and chocolate will be served in room 04.3.15 after the seminar.