DAILY PAPER REVIEW

0209_A fully robust PARAFAC method for analyzing fluorescence data

 

 

ESEL Paper Review_20110126 by Aamir Alaud-din

1.    Title and Author

Title: 2011_02_09_A fully robust PARAFAC method for analyzing fluorescence data

Authors: Sanne Engelena, Stina Froschb* and Bo Munk Jørgensenb

aKatholieke Universiteit Leuven, Department of Mathematics, W. De Croylaan 54, 3001 Leuven, Belgium

b*Correspondence to: S. Frosch, National Institute of Aquatic Resources, Technical University of Denmark, Søltofts Plads, Building 221 Lyngby 2800, Denmark.

bDTU Aqua, National Institute of Aquatic Resources, Søltofts Plads, Building 221, 2800 Kgs. Lyngby, Denmark


2.    Summary of Paper

In matrix notations, parallel factor analysis (PARAFAC), which is one of the commonly used techniques to model excitation-emission data, can be written as:

 

Irregular EEM landscapes are called the outliers and need to be removed from the data. PARAFAC is based on least square technique. So, if outliers are not removed from the data set, the model will explain the outliers instead of majority of the samples.

Another important issue in PARAFAC modeling is to deal with first and second order Rayleigh and Raman scatters. It is because these scatters do not contain any chemical information.

The paper deals with the combined approach of PARAFAC modeling with automatic scatter identification.

For 10 generated data sets, mean squared error was computed as follows:

 

with  and   where   is an outlier, otherwise  .

Percentage variance explained was estimated by:

 

with  

The technique was tested for
a.    Clean data
b.    Data with scatter and without outliers
c.    Data with outliers and without scatter
d.    Data with outlier and scatter
The combined PARAFAC algorithm was finally used on the freeware laboratory-made data set. The data consisted of 27 samples of four different fluorophores. The excitation wavelength was in the range of 200 ? 315nm. The range of emission wavelength was from 250 ? 482nm.
Previous papers discussed one aspect at one time, either outliers were handled or the scatters were handled. The combined algorithm handled with scatters and outliers simultaneously. This clearly showed that the combined approach is better in performance than the previous techniques and algorithms for PARAFAC modeling of excitation ? emission data.
3.    Contribution to ESEL

This paper gives a way to handle the scatter and outlying samples simultaneously. So, a try to incorporate different features at one place is time demanding but it, sometimes, is possible for better results.

By: Aamir Alaud-din
aamiralauddin@gist.ac.kr

첨부 (1)
20110209.pdf
259.8KB / Download 1