Wyniki wyszukiwania - BazTech

Ograniczanie wyników

Znaleziono wyników: 1

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: spectra pre-processing

Sortuj według:

Ogranicz wyniki do:

Evaluation of available protein mass spectra pre-processing algorithms

Marczyk M.

Bio-Algorithms and Med-Systems

2011

Vol. 7, no. 2

25--30

Analysis of the profiles of body fluids by MALDI – ToF mass spectrometry is a popular technique used for searching of protein biomarkers that have potential application in early detection and diagnose of a cancer. A typical strategy used in validation of new biomarkers involves two basic aspects. Predictive properties of discovered differing features must be proven by classification of patients and controls spectra. Then, these features must be connected to proteins / peptides present in the analysed samples. Therefore, most mass spectra pre-processing procedures are based on reducing dimensionality of the spectrum to only significant features (peaks), assuming that each peak corresponds to a single protein / peptide and its position in the m/z scale, and height carry direct information on the composition of the test substance. In the literature there are several different approaches of mass spectra pre-processing. But so far there are no standards for selection of techniques that are the most effective for this type of data. Only the pre-processing steps that should be done in order to extract the desired information from raw spectra are specified. This paper presents some algorithms that are compared on two levels. By the classification of real data sets differentiating potential of detected features was examined and the ability to reconstruct proteins / peptides in the test sample was checked. Since the composition of the specimen is not known, we used a virtual machine to generate artificial spectra. Despite many published studies, scientists searching for biomarkers, still encounter many serious problems. Existing methods for mass spectra pre-processing are very sensitive to changes in data collection protocols, or instrumentation. Identified biomarkers of cancer vary between different research groups. The key is to choose the appropriate settings of the methods used. Thus, there is a need to test new procedures and automate the tuning of parameters of existing algorithms. Our simulations showed, that Align and CWT algorithms eliminates false positive peaks efficiently and that Align is the most flexible for changes in signal quality from all studied mass spectra pre-processing packages.