Abstract
Bioinformatics has become an important tool to support clinical and biological research and the analysis of functional data, is a common task in bioinformatics (Schleif, 2006). Gene analysis in form of micro array analysis (Schena, 1995) and protein analysis (Twyman, 2004) are the most important fields leading to multiple sub omics-disciplines like pharmacogenomics, glycoproteomics or metabolomics. Measurements of such studies are high dimensional functional data with few samples for specific problems (Pusch, 2005). This leads to new challenges in the data analysis. Spectra of mass spectrometric measurements are such functional data requiring an appropriate analysis (Schleif, 2006). Here we focus on the determination of classification models for such data. In general, the spectra are transformed into a vector space followed by training a classifier (Haykin, 1999). Hereby the functional nature of the data is typically lost. We present a method which takes this specific data aspects into account. A wavelet encoding (Mallat, 1999) is applied onto the spectral data leading to a compact functional representation. Subsequently the Supervised Neural Gas classifier (Hammer, 2005) is applied, capable to handle functional metrics as introduced by Lee & Verleysen (Lee, 2005). This allows the classifier to utilize the functional nature of the data in the modelling process. The presented method is applied to clinical proteome data showing good results and can be used as a bioinformatics method for biomarker discovery.
TopBackground
Applications of mass spectrometry (ms) in clinical proteomics have gained tremendous visibility in the scientific and clinical community (Villanueva, 2004) (Ketterlinus, 2005). One major objective is the search for potential classification models for cancer studies, with strong requirements for validated signal patterns (Ransohoff, 2005). Primal optimistic results as given in (Petricoin, 2002) are now considered more carefully, because the complexity of the task of biomarker discovery and an appropriate data processing has been observed to be more challenging than expected (Ransohoff, 2005). Consequently the main recent work in this field is focusing on optimization and standardisation. This includes the biochemical part (e.g. Baumann, 2005), the measurement (Orchard, 2003) and the subsequently data analysis (Morris, 2005)(Schleif 2006).
Key Terms in this Chapter
Bioinformatics: Generic term of a research field as well as a set of methods used in computational biology or medicine to analyse multiple kinds of biological or clinical data. It combines the disciplines of computer science, artificial intelligence, applied mathematics, statistics, biology, chemistry and engineering in the field of biology and medicine. Typical research subjects are problem adequate data pre-processing of measured biological sample information (e.g. data cleaning, alignments, feature extraction), supervised and unsupervised data analysis (e.g. classification models, visualization, clustering, biomarker discovery) and multiple kinds of modelling (e.g. protein structure prediction, analysis of expression of gene, proteins, gene/protein regulation networks/interactions) for one or multidimensional data including time series. Thereby the most common problem is the high dimensionality of the data and the small number of samples which in general make standard approach (e.g. classical statistic) inapplicable.
Wavelet Analysis: Method used in signal processing to analyse a signal by means of frequency and local information. Thereby the signal is encoded in a representation of wavelets, which are specific kinds of mathematical functions. The Wavelet encoding allows the representation of the signal at different resolutions, the coefficients contain frequency information but can also be localized in the signal.
Clinical Proteomics: Proteomics is the field of research related to the analysis of the proteome of an organism. Thereby, clinical proteomics is focused on research mainly related to disease prediction and prognosis in the clinical domain by means of proteome analysis. Standard methods for proteome analysis are available by Mass spectrometry.
Relevance Learning: A method, typically used in supervised classification, to determine problem specific metric parameter. With respect to the used metric and learning schema univariate, correlative and multivariate relations between data dimensions can be analyzed. Relevance learning typically leads to significantly improved, problem adapted metric parameters and classification models.
Mass Spectrometry: An analytical technique used to measure the mass-to-charge ratio of ions. In clinical proteomics mass spectrometry can be applied to extract fingerprints of samples (like blood, urine, bacterial extracts) whereby semi-quantitative intensity differences between sample cohorts may indicate biomarker candidates
Biomarker: Mainly in clinical research one goal of experiments is to determine patterns which are predictive for the presents or prognosis of a disease state, frequently called biomarker. Biomarkers can be single or complex (pattern) indicator variables taken from multiple measurements of a sample. The ideal biomarker has a high sensitivity, specificity and is reproducible (under standardized conditions) with respect to control experiments in other labs. Further it can be expected that the marker is vanishing or changing during a treatment of the disease.
Prototype Classifiers: Are a specific kind of neural networks and related to the kNN classifier. The classification model consists of so called prototypes which are representatives for a larger set of data points. The classification is done by a nearest neighbour classification using the prototypes. Nowadays prototype classifiers can be found in multiple fields (robotics, character recognition, signal processing or medical diagnosis) trained to find (non)linear relationships in data.