ClaNC: point-and-click software for classifying microarrays to nearest centroids

SUMMARY: ClaNC (classification to nearest centroids) is a simple and an accurate method for classifying microarrays. This document introduces a point-and-click interface to the ClaNC methodology. The software is available as an R package. AVAILABILITY: ClaNC is freely available from http://students.washington.edu/adabney/clanc     

Improved centroids estimation for the nearest shrunken centroid classifier

MOTIVATION: The nearest shrunken centroid (NSC) method has been successfully applied in many DNA-microarray classification problems. The NSC uses 'shrunken' centroids as prototypes for each class and identifies subsets of genes that best characterize each class. Classification is then made to the nearest (shrunken) centroid. The NSC is very easy to implement and very easy to interpret, however, it has drawbacks. RESULTS: We show that the NSC method can be interpreted in the framework of LASSO regression. Based on that, we consider two new methods, adaptive L(infinity)-norm penalized NSC (ALP-NSC) and adaptive hierarchically penalized NSC (AHP-NSC), with two different penalty functions for microarray classification, which improve over the NSC. Unlike the L(1)-norm penalty used in LASSO, the penalty terms that we consider make use of the fact that parameters belonging to one gene should be treated as a natural group. Numerical results indicate that the two new methods tend to remove irrelevant genes more effectively and provide better classification results than the L(1)-norm approach. AVAILABILITY: R code for the ALP-NSC and the AHP-NSC algorithms are available from authors upon request.     

Classification of gene microarrays by penalized logistic regression

Classification of patient samples is an important aspect of cancer diagnosis and treatment. The support vector machine (SVM) has been successfully applied to microarray cancer diagnosis problems. However, one weakness of the SVM is that given a tumor sample, it only predicts a cancer class label but does not provide any estimate of the underlying probability. We propose penalized logistic regression (PLR) as an alternative to the SVM for the microarray cancer diagnosis problem. We show that when using the same set of genes, PLR and the SVM perform similarly in cancer classification, but PLR has the advantage of additionally providing an estimate of the underlying probability. Often a primary goal in microarray cancer diagnosis is to identify the genes responsible for the classification, rather than class prediction. We consider two gene selection methods in this paper, univariate ranking (UR) and recursive feature elimination (RFE). Empirical results indicate that PLR combined with RFE tends to select fewer genes than other methods and also performs well in both cross-validation and test samples. A fast algorithm for solving PLR is also described.     

The effects of mismatches on hybridization in DNA microarrays: determination of nearest neighbor parameters

Quantifying interactions in DNA microarrays is of central importance for a better understanding of their functioning. Hybridization thermodynamics for nucleic acid strands in aqueous solution can be described by the so-called nearest neighbor model, which estimates the hybridization free energy of a given sequence as a sum of dinucleotide terms. Compared with its solution counterparts, hybridization in DNA microarrays may be hindered due to the presence of a solid surface and of a high density of DNA strands. We present here a study aimed at the determination of hybridization free energies in DNA microarrays. Experiments are performed on custom Agilent slides. The solution contains a single oligonucleotide. The microarray contains spots with a perfect matching (PM) complementary sequence and other spots with one or two mismatches (MM) : in total 1006 different probe spots, each replicated 15 times per microarray. The free energy parameters are directly fitted from microarray data. The experiments demonstrate a clear correlation between hybridization free energies in the microarray and in solution. The experiments are fully consistent with the Langmuir model at low intensities, but show a clear deviation at intermediate (non-saturating) intensities. These results provide new interesting insights for the quantification of molecular interactions in DNA microarrays.     

Classification of protein profiles from antibody microarrays using heat and detergent treatment

Antibody microarrays offer new opportunities for exploring the proteome and to identify biomarker candidates in human serum and plasma. Here, we have investigated the effect of heat and detergents on an antibody-based suspension bead array (SBA) assay using polyclonal antibodies and biotinylated plasma samples. With protein profiles from more than 2300 antibodies generated in 384-plex antibody SBAs, three major classes of heat and detergent susceptibility could be described. The results show that washing of the beads with SDS (rather than Tween) after target binding lowered intensity levels of basically all profiles and that about 50% of the profiles appeared to be lowered to a similar extent by heating of the sample. About 33% of the profiles appeared to be insensitive to heat treatment while another 17% showed a positive influence of heat to yield elevated profiles. The results suggest that the classification of antibodies is driven by the molecular properties of the antibody-antigen interaction and can generally not be predicted based on protein class or Western blot data. The experimental scheme presented here can be used to systematically categorize antibodies and thereby combine antibodies with similar properties into targeted arrays for analysis of plasma and serum.     

From microarrays to genome duplications

A report on the fifth annual conference of the Society for Bioinformatics in the Nordic Countries (SOCBIN), 'Bioinformatics 2003', Helsinki, Finland, 22-24 May 2003.     

Intact cell adhesion to glycan microarrays

A rapid and reproducible method was developed to detect and quantify carbohydrate-mediated cell adhesion to glycans arrayed on glass slides. Monosaccharides and oligosaccharides were covalently attached to glass slides in 1.7-mm-diameter spots (200 spots/slide) separated by a Teflon gasket. Primary chicken hepatocytes, which constitutively express a C-type lectin that binds to nonreducing terminal N-acetylglucosamine residues, were labeled with a fluorescent dye and incubated in 1.3-microL aliquots on the glycosylated spots. After incubating to allow cell adhesion, nonadherent cells were removed by immersing the slide in phosphate buffered saline, inverting, and centrifuging in a sealed custom acrylic chamber so that cells on the derivatized spots were subjected to a uniform and controlled centrifugal detachment force while avoiding an air-liquid interface. After centrifugation, adherent cells were fixed in place and detected by fluorescent imaging. Chicken hepatocytes bound to nonreducing terminal GlcNAc residues in different linkages and orientations but not to nonreducing terminal galactose or N-acetylgalactosamine residues. Addition of soluble GlcNAc (but not Gal) prior to incubation reduced cell adhesion to background levels. Extension of the method to CD4+ human T-cells on a 45-glycan diversity array revealed specific adhesion to the sialyl Lewis x structure. The described method is a robust approach to quantify selective cell adhesion using a wide variety of glycans and may contribute to the repertoire of tools for the study of glycomics.     

Application of independent component analysis to microarrays

We apply linear and nonlinear independent component analysis (ICA) to project microarray data into statistically independent components that correspond to putative biological processes, and to cluster genes according to over- or under-expression in each component. We test the statistical significance of enrichment of gene annotations within clusters. ICA outperforms other leading methods, such as principal component analysis, k-means clustering and the Plaid model, in constructing functionally coherent clusters on microarray datasets from Saccharomyces cerevisiae, Caenorhabditis elegans and human.     

Application of phenotypic microarrays to environmental microbiology

Environmental organisms are extremely diverse and only a small fraction has been successfully cultured in the laboratory. Culture in micro wells provides a method for rapid screening of a wide variety of growth conditions and commercially available plates contain a large number of substrates, nutrient sources, and inhibitors, which can provide an assessment of the phenotype of an organism. This review describes applications of phenotype arrays to anaerobic and thermophilic microorganisms, use of the plates in stress response studies, in development of culture media for newly discovered strains, and for assessment of phenotype of environmental communities. Also discussed are considerations and challenges in data interpretation and visualization, including data normalization, statistics, and curve fitting.     

Using DNA microarrays to study natural variation

The emerging field of genomics examines the relationship between genetic and phenotypic variation by describing and analyzing patterns of natural variation on a genome-wide scale. In this endeavor, an important tool is the use of microarrays, which enable simultaneous screening of thousands of assays. Microarrays were originally designed for the detection of differences between samples and are thus ideally suited to high-throughput studies of natural variation. Novel microarray platforms enable the high throughput survey of variation at multiple levels, including DNA sequences, gene expression, protein binding, and methylation. However, most microarray data analysis tools, notably normalization methods, were developed for experiments in which only few features differed between samples. In studies of natural variation, this assumption does not always hold, raising a number of new challenges.     

Application of sector protein microarrays to clinical samples

Many protein functions are conferred by posttranslational modifications, which allow proteins to perform specific cellular tasks. Protein microarrays enable specific detection of posttranslational modifications not attainable by gene arrays. Reverse-phase protein microarrays have been widely adopted for use with clinical biopsy specimens because they have many advantages including highly reproducible printing of cellular lysates onto array surfaces, buit-in dilution curves, and direct detection using one antibody per analyte. This results in high-sensitivity, broad dynamic range, and favorable precision. Reverse-phase arrays have been restricted to a one slide/one antibody format. Although this is suitable for analyzing treatment effects over populations of samples, it is not well suited to individual patient assessments. One means of reaching this goal is the sector array format. Through the sector array, multiple antibody probes can be multiplexed on a single slide containing replicate immobilized aliquots from one patient. Thus, on one slide, a complete set of analytes can be characterized and used to support a therapy decision. This article describes a method for constructing sector arrays and demonstrates feasibility and adequate sensitivity applied to apoptosis related pathways.     

Learning to fly--getting the best out of microarrays

Microarrays are becoming pervasive in biology and can offer unparalleled insights into biological processes. This article suggests that a more considered approach to the design of array probe sequences can substantially improve the technology.     

Application of DNA microarrays to study human alcoholism

An emerging idea is that long-term alcohol abuse results in changes in gene expression in the brain and that these changes are responsible at least partly for alcohol tolerance, dependence and neurotoxicity. The overall goal of our research is to identify genes which are differentially expressed in the brains of well-characterized human alcoholics as compared with non-alcoholics. This should identify as-yet-unknown alcohol-responsive genes, and may well confirm changes in the expression of genes which have been delineated in animal models of alcohol abuse. Cases were carefully selected and samples pooled on the basis of relevant criteria; differential expression was monitored by microarray hybridization. The inherent diversity of human alcoholics can be exploited to identify genes associated with specific pathological processes, as well as to assess the effects of concomitant disease, severity of brain damage, drinking behavior, and factors such as gender and smoking history. Initial results show selective changes in gene expression in alcoholics; of particular importance is a coordinated reduction in genes coding for myelin components.     

Genome analysis with distance to the nearest dissimilar nucleotide

DNA may be represented by sequences of four symbols, but it is often useful to convert those symbols into real or complex numbers for further analysis. Several mapping schemes have been used in the past, but most of them seem to be unrelated to any intrinsic characteristic of DNA. The objective of this work was to study a mapping scheme that is directly related to DNA characteristics, and that could be useful in discriminating between different species.Recently, we have proposed a methodology based on the inter-nucleotide distance, which proved to contribute to the discrimination among species. In this paper, we introduce a new distance, the distance to the nearest dissimilar nucleotide, which is the distance of a nucleotide to first occurrence of a different nucleotide. This distance is related to the repetition structure of single nucleotides. Using the information resulting from the concatenation of the distance to the nearest dissimilar and the inter-nucleotide distance, we found that this new distance brings additional discriminative capabilities. This suggests that the distance to the nearest dissimilar nucleotide might contribute with useful information about the evolution of the species.     

Challenges in applying microarrays to environmental studies

Although DNA microarray technology has been used successfully to analyze global gene expression in pure cultures, it has not been rigorously tested and evaluated within the context of complex environmental samples. Adapting microarray hybridization for use in environmental studies faces several challenges associated with specificity, sensitivity and quantitation.     

Tissue microarrays

The identification of disease-related genes is a major focus of modern biomedical research. Recent techniques, including array-based platforms for molecular profiling of disease tissues such as DNA arrays for expression profiling or matrix comparative genomic hybridization, allow for the comprehensive screening of the whole genome in a single experiment. Consequently, thousands of candidate genes have already been identified that may be linked to disease development and progression, and the process of lead discovery continues unimpeded. The evaluation of the clinical value of such leads is challenging because thousands of well-characterized tissue specimens must be analyzed. Tissue microarray (TMA) technology enables high-throughput tissue analyses to keep pace with the rapid process of lead discovery. With this technique, up to 1000 minute tissue samples are brought into an array format and analyzed simultaneously. The TMA technology is a fast, cost-effective, and statistically powerful method that will substantially facilitate translational research.     

Carbohydrate microarrays

Sugar chains are abundantly expressed on the outer surfaces of the vast majority of viral, bacterial, protozoan and fungal pathogens, as well as on the membranes of mammalian cells. This class of carbohydrate molecule is without peer in structural diversity and is characteristically suitable for storing and displaying biological signals for molecular and cellular recognition. Exploring the biological information contained in sugar chains is an important topic of current postgenomic research. To facilitate these investigations, we have focused on the establishment of a carbohydrate-based microarray technology. Recently, we reported that a large panel of carbohydrate-containing macromolecules, including polysaccharides, natural glycoconjugates, and the mono- and oligosaccharides coupled to carrier molecules, can be stably immobilized on a microglass slide to produce a large-scale carbohydrate microarray. In this review, we attempt to summarize our recent progress in using this technology to uncover the carbohydrate-based biological signals that are recognized by the human and animal immune systems. We also discuss the potential of various platforms of carbohydrate microarrays that were recently established and analyze the challenges to future development of carbohydrate microarray technologies and their applications.