Grzegorz M. Boratyn

Urinary Peptidome May Predict Renal Function Decline in Type 1 Diabetes and Microalbuminuria

One third of patients with type 1 diabetes and microalbuminuria experience an early, progressive decline in renal function that leads to advanced stages of chronic kidney disease and ESRD. We hypothesized that the urinary proteome may distinguish between stable renal function and early renal function decline among patients with type 1 diabetes and microalbuminuria. We followed patients with normal renal function and microalbuminuria for 10 to 12 yr and classified them into case patients (n = 21) with progressive early renal function decline and control subjects (n = 40) with stable renal function. Using liquid chromatography matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, we identified three peptides that decreased in the urine of patients with early renal function decline [fragments of alpha1(IV) and alpha1(V) collagens and tenascin-X] and three peptides that increased (fragments of inositol pentakisphosphate 2-kinase, zona occludens 3, and FAT tumor suppressor 2). In renal biopsies from patients with early nephropathy from type 1 diabetes, we observed increased expression of inositol pentakisphosphate 2-kinase, which was present in granule-like cytoplasmic structures, and zona occludens 3. These results indicate that urinary peptide fragments reflect changes in expression of intact protein in the kidney, suggesting new potential mediators of diabetic nephropathy and candidate biomarkers for progressive renal function decline.

Hybridization of independent component analysis, rough sets, and multi-objective evolutionary algorithms for classificatory decomposition of cortical evoked potentials

This article presents a continuation of our research aiming at improving the effectiveness of signal decomposition algorithms by providing them with “classification-awareness.” We investigate hybridization of multi-objective evolutionary algorithms (MOEA) and rough sets (RS) to perform the task of decomposition in the light of the underlying classification problem itself. In this part of the study, we also investigate the idea of utilizing the Independent Component Analysis (ICA) to initialize the population in the MOEA.

Incorporation of biological knowledge into distance for clustering genes.

In this paper we propose a data based algorithm to marry existing biological knowledge (e.g., functional annotations of genes) with experimental data (gene expression profiles) in creating an overall dissimilarity that can be used with any clustering algorithm that uses a general dissimilarity matrix. We explore this idea with two publicly available gene expression data sets and functional annotations where the results are compared with the clustering results that uses only the experimental data. Although more elaborate evaluations might be called for, the present paper makes a strong case for utilizing existing biological information in the clustering process. Availability Supplement is available at www.somnathdatta.org/Supp/Bioinformation/appendix.pdf

Evolutionary Algorithms and Rough Sets-Based Hybrid Approach to Classificatory Decomposition of Cortical Evoked Potentials

This paper presents a novel approach to decomposition and classification of rats cortical evoked potentials (EPs). The decomposition is based on learning of a sparse set of basis functions using Evolutionary Algorithms (EAs). The basis functions are generated in a potentially overcomplete dictionary of the EP components according to a probabilistic model of the data. Compared to the traditional, statistical signal decomposition techniques, this allows for a number of basis functions greater than the dimensionality of the input signals, which can be of a great advantage. However, there arises an issue of selecting the most significant components from the possibly overcomplete collection. This is especially important in classification problems performed on the decomposed representation of the data, where only those components that provide a substantial discernibility between EPs of different groups are relevant. In this paper, we propose an approach based on the Rough Set theorys (RS) feature selection mechanisms to deal with this problem. We design an EA and RS-based hybrid system capable of signal decomposition and, based on a reduced component set, signal classification.

Sparse Correlation Kernel Analysis and Evolutionary Algorithm-Based Modeling of the Sensory Activity within the Rat's Barrel Cortex

This paper presents a new paradigm for signal decomposition and reconstruction that is based on the selection of a sparse set of basis functions. Based on recently reported results, we note that this framework is equivalent to approximating the signal using Support Vector Machines. Two different algorithms of modeling sensory activity within the barrel cortex of a rat are presented. First, a slightly modified approach to the Independent Component Analysis (ICA) algorithm and its application to the investigation of Evoked Potentials (EP), and second, an Evolutionary Algorithm (EA) for learning an overcomplete basis of the EP components by viewing it as probabilistic model of the observed data. The results of the experiments conducted using these two approaches as well as a discussion concerning a possible utilization of those results are also provided.

Biologically Supervised Hierarchical Clustering Algorithms for Gene Expression Data

Cluster analysis has become a standard part of gene expression analysis. In this paper, we propose a novel semi-supervised approach that offers the same flexibility as that of a hierarchical clustering. Yet it utilizes, along with the experimental gene expression data, common biological information about different genes that is being complied at various public, Web accessible databases. We argue that such an approach is inherently superior than the standard unsupervised approach of grouping genes based on expression data alone. It is shown that our biologically supervised methods produce better clustering results than the corresponding unsupervised methods as judged by the distance from the model temporal profiles. R-codes of the clustering algorithm are available from the authors upon request

Hybridization of Blind Source Separation and Rough Sets for Proteomic Biomarker Indentification

Biomarkers are molecular parameters associated with presence and severity of specific disease states. Search for biological markers of cancer in proteomic profiles is a relatively new but very active research area. This paper presents a novel approach to feature selection and thus biomarker identification. The proposed method is based on blind separation of sources and selection of features from a reduced set of components.

Utlization of Human Expert Techniques for Detection of Low-Abundant Peaks in High-Resolution Mass Spectra

Interpretation and classification of mass spectra is usually performed using a list of peaks as their mathematical representation. This fact makes peak detection a bottleneck of mass spectra analysis, since quality and biological relevance of any results strongly depends on the accuracy of peak detection process. Many algorithms utilize intensity to differentiate between peaks and noise and thus bias the detection process to the high abundant peaks. However important information may also be contained in the lower-intensity peaks that are more difficult to discover. We present an algorithm specifically designed for detection of low-abundant peaks

Sparse coding and rough set theory-based hybrid approach to the classificatory decomposition of cortical evoked potentials

This paper presents a novel approach to classification of decomposed cortical evoked potentials (EPs). The decomposition is based on learning of a sparse set of basis functions using an artificial neural network (ANN). The basis functions are generated according to a probabilistic model of the data. In contrast to the traditional signal decomposition techniques (i.e. principle component analysis or independent component analysis), this allows for an overcomplete representation of the data (i.e. number of basis functions that is greater than the dimensionality of the input signals). Obviously, this can be of a great advantage. However, there arises an issue of selecting the most significant components from the whole collection. This is especially important in classification problems based upon the decomposed representation of the data, where only those components that provide a substantial discernibility between EPs of different groups are relevant. To deal with this problem, we propose an approach based on the rough set theorys (RS) feature selection mechanisms. We design a sparse coding- and RS-based hybrid system capable of signal decomposition and, based on a reduced component set, signal classification.

Bayesian approach to analysis of protein patterns for identification of myeloma cancer

Early detection is critical in cancer control and prevention. Proteomics is an area in discovery of biomarkers that are molecular parameters associated with presence and severity of specific disease states. Protein samples are analyzed on the basis of mass to charge ratio (m/z) of particles they are composed of. Sequences of intensities (i.e. number of particles with specific value of m/z) can be interpreted using statistical approaches or information theory and data mining tools. The data mining, statistical, and information theoretical approaches have already been successfully applied to identify several types of cancer in gene or protein samples. However, due to small size of training sets and very large dimensionality of this kind of data new approaches that incorporate theoretical information need to be developed. This paper presents an application of Bayesian approach to detection of myeloma cancer sites in a sequence of ion intensity values obtained from protein samples. The use of scoring function developed by authors for calculation of data likelihood is also proposed.