BACKGROUND: Sporadic Amyotrophic Lateral Sclerosis (sALS) is a devastating, complex disease of unknown etiology. We studied this disease with microarray technology to capture as much biological complexity as possible. The Affymetrix-focused BaFL pipeline takes into account problems with probes that arise from physical and biological properties, so we adapted it to handle the long-oligonucleotide probes on our arrays (hence LO-BaFL). The revised method was tested against a validated array experiment and then used in a meta-analysis of peripheral white blood cells from healthy control samples in two experiments. We predicted differentially expressed (DE) genes in our sALS data, combining the results obtained using the TM4 suite of tools with those from the LO-BaFL method. Those predictions were tested using qRT-PCR assays. RESULTS: LO-BaFL filtering and DE testing accurately predicted previously validated DE genes in a published experiment on coronary artery disease (CAD). Filtering healthy control data from the sALS and CAD studies with LO-BaFL resulted in highly correlated expression levels across many genes. After bioinformatics analysis, twelve genes from the sALS DE gene list were selected for independent testing using qRT-PCR assays. High-quality RNA from six healthy Control and six sALS samples yielded the predicted differential expression for 7 genes: TARDBP, SKIV2L2, C12orf35, DYNLT1, ACTG1, B2M, and ILKAP. Four of the seven have been previously described in sALS studies, while ACTG1, B2M and ILKAP appear in the context of this disease for the first time. Supplementary material can be accessed at: http://webpages.uncc.edu/~cbaciu/LO-BaFL/supplementary_data.html. CONCLUSION: LO-BaFL predicts DE results that are broadly similar to those of other methods. The small healthy control cohort in the sALS study is a reasonable foundation for predicting DE genes. Modifying the BaFL pipeline allowed us to remove noise and systematic errors, improving the power of this study, which had a small sample size. Each bioinformatics approach revealed DE genes not predicted by the other; subsequent PCR assays confirmed seven of twelve candidates, a relatively high success rate.

BACKGROUND: Probes on a microarray represent a frozen view of a genome and are quickly outdated when new sequencing studies extend our knowledge, resulting in significant measurement error when analyzing any microarray experiment. There are several bioinformatics approaches to improve probe assignments, but without in-house programming expertise, standardizing these custom array specifications as a usable file (e.g. as Affymetrix CDFs) is difficult, owing mostly to the complexity of the specification file format. However, without correctly standardized files there is a significant barrier for testing competing analysis approaches since this file is one of the required inputs for many commonly used algorithms. The need to test combinations of probe assignments and analysis algorithms led us to develop ArrayInitiative, a tool for creating and managing custom array specifications. RESULTS: ArrayInitiative is a standalone, cross-platform, rich client desktop application for creating correctly formatted, custom versions of manufacturer-provided (default) array specifications, requiring only minimal knowledge of the array specification rules and file formats. Users can import default array specifications, import probe sequences for a default array specification, design and import a custom array specification, export any array specification to multiple output formats, export the probe sequences for any array specification and browse high-level information about the microarray, such as version and number of probes. The initial release of ArrayInitiative supports the Affymetrix 3' IVT expression arrays we currently analyze, but as an open source application, we hope that others will contribute modules for other platforms. CONCLUSIONS: ArrayInitiative allows researchers to create new array specifications, in a standard format, based upon their own requirements. This makes it easier to test competing design and analysis strategies that depend on probe definitions. Since the custom array specifications are easily exported to the manufacturer's standard format, researchers can analyze these customized microarray experiments using established software tools, such as those available in Bioconductor.

BACKGROUND: Microarrays depend on appropriate probe design to deliver the promise of accurate genome-wide measurement. Probe design, ideally, produces a unique probe-target match with homogeneous duplex stability over the complete set of probes. Much of microarray pre-processing is concerned with adjusting for non-ideal probes that do not report target concentration accurately. Cross-hybridizing probes (non-unique), probe composition and structure, as well as platform effects such as instrument limitations, have been shown to affect the interpretation of signal. Data cleansing pipelines seldom filter specifically for these constraints, relying instead on general statistical tests to remove the most variable probes from the samples in a study. This adjusts probes contributing to ProbeSet (gene) values in a study-specific manner. We refer to the complete set of factors as biologically applied filter levels (BaFL) and have assembled an analysis pipeline for managing them consistently. The pipeline and associated experiments reported here examine the outcome of comprehensively excluding probes affected by known factors on inter-experiment target behavior consistency. RESULTS: We present here a 'white box' probe filtering and intensity transformation protocol that incorporates currently understood factors affecting probe and target interactions; the method has been tested on data from the Affymetrix human GeneChip HG-U95Av2, using two independent datasets from studies of a complex lung adenocarcinoma phenotype. The protocol incorporates probe-specific effects from SNPs, cross-hybridization and low heteroduplex affinity, as well as effects from scanner sensitivity, sample batches, and includes simple statistical tests for identifying unresolved biological factors leading to sample variability. Subsequent to filtering for these factors, the consistency and reliability of the remaining measurements is shown to be markedly improved. CONCLUSIONS: The data cleansing protocol yields reproducible estimates of a given probe or ProbeSet's (gene's) relative expression that translates across datasets, allowing for credible cross-experiment comparisons. We provide supporting evidence for the validity of removing several large classes of probes, and for our approaches for removing outlying samples. The resulting expression profiles demonstrate consistency across the two independent datasets. Finally, we demonstrate that, given an appropriate sampling pool, the method enhances the t-test's statistical power to discriminate significantly different means over sample classes.

BACKGROUND: Affymetrix gene expression arrays incorporate paired perfect match (PM) and mismatch (MM) probes to distinguish true signals from those arising from cross-hybridization events. A MM signal often shows greater intensity than a PM signal; we propose that one underlying cause is the presence of allelic variants arising from single nucleotide polymorphisms (SNPs). To annotate and characterize SNP contributions to anomalous probe binding behavior we have developed a software tool called AffyMAPSDetector. RESULTS: AffyMAPSDetector can be used to describe any Affymetrix expression GeneChip with respect to SNPs. When AffyMAPSDetector was run on GeneChip HG-U95Av2 against dbSNP-build-123, we found 7286 probes (belonging to 2,582 probesets) containing SNPs, out of which 325 probes contained at least one SNP at position 13. Against dbSNP-build-126, 8758 probes (belonging to 3,002 probesets) contained SNPs, of which 409 probes contained at least one SNP at position 13. Therefore, depending on the expressed allele, the MM probe can sometimes be the transcript complement. This information was used to characterize probe measurements reported in a published, well-replicated lung adenocarcinoma study. The total intensity distributions showed that the SNP-containing probes had a larger negative mean intensity difference (PM-MM) and greater range of the difference than did probes without SNPs. In the sample replicates, SNP-containing probes with reproducible intensity ratios were identified, allowing selection of SNP probesets that yielded unique sample signatures. At the gene expression level, use of the (MM-PM) value for SNP-containing probes resulted in different Presence/Absence calls for some genes. Such a change in status of the genes has the clear potential for influencing downstream clustering and classification results. CONCLUSION: Output from this tool characterizes SNP-containing probes on GeneChip microarrays, thus improving our understanding of factors contributing to expression measurements. The pattern of SNP binding examined so far indicates distinct behavior of the SNP-containing probes and has the potential to help us identify new SNPs. Knowing which probes contain SNPs provides flexibility in determining whether to include or exclude them from gene-expression intensity calculations; selected sets of SNP-containing probes produce sample-unique signatures. AffyMAPSDetector information is available at http://www.binf.gmu.edu/weller/BMC_bioinformatics/AffyMapsDetector/index.html.

BACKGROUND: Accurate methods for extraction of meaningful patterns in high dimensional data have become increasingly important with the recent generation of data types containing measurements across thousands of variables. Principal components analysis (PCA) is a linear dimensionality reduction (DR) method that is unsupervised in that it relies only on the data; projections are calculated in Euclidean or a similar linear space and do not use tuning parameters for optimizing the fit to the data. However, relationships within sets of nonlinear data types, such as biological networks or images, are frequently mis-rendered into a low dimensional space by linear methods. Nonlinear methods, in contrast, attempt to model important aspects of the underlying data structure, often requiring parameter(s) fitting to the data type of interest. In many cases, the optimal parameter values vary when different classification algorithms are applied on the same rendered subspace, making the results of such methods highly dependent upon the type of classifier implemented. RESULTS: We present the results of applying the spectral method of Lafon, a nonlinear DR method based on the weighted graph Laplacian, that minimizes the requirements for such parameter optimization for two biological data types. We demonstrate that it is successful in determining implicit ordering of brain slice image data and in classifying separate species in microarray data, as compared to two conventional linear methods and three nonlinear methods (one of which is an alternative spectral method). This spectral implementation is shown to provide more meaningful information, by preserving important relationships, than the methods of DR presented for comparison. Tuning parameter fitting is simple and is a general, rather than data type or experiment specific approach, for the two datasets analyzed here. Tuning parameter optimization is minimized in the DR step to each subsequent classification method, enabling the possibility of valid cross-experiment comparisons. CONCLUSION: Results from the spectral method presented here exhibit the desirable properties of preserving meaningful nonlinear relationships in lower dimensional space and requiring minimal parameter fitting, providing a useful algorithm for purposes of visualization and classification across diverse datasets, a common challenge in systems biology.

GenoMosaic is a portable database application for on demand multiple genome comparison. We discuss the methods used to generate a GenoMosaic data set from genome sequence data, and present the relational data model used in the application. We define an abstraction of genome sequence data (the feature mosaic) that allows us to bridge between annotation that describes features within single genes and that which includes possibly multiple genes and intergenic features over long stretches of genomic sequence. The goal of this project is to support new method development for on-demand multiple genome comparison. Each genome to be compared can be modeled as a string of generic features of any type that can be computationally defined, related by adjacency information within and among genomes. The generic feature abstraction makes it possible to study the arrangement of features in the genome at a level of detail which includes RNA genes, putative regulatory regions, SNPs, overlapping transcripts, intron splice junctions, alternative polyadenylation signals-in short, to incorporate significant sequence details which are not necessarily within protein-coding regions. This abstraction is amenable to functional implementation as a relational data model upon which novel query capabilities can be built, and provides objects that can be analyzed using algorithms for comparison of strings and lists. As an initial effort, we have implemented a prototype using a representative set of comparative and content-based annotation methods to reduce a collection of prokaryotic genomes to a feature mosaic representation. Entity-Relationship modeling was then used to develop a data model capable of storing detailed results, including complete parameters for each instance of analysis.

The EST Analysis Pipeline (ESTAP) is a set of analytical procedures that automatically verify, cleanse, store and analyze ESTs generated on high-throughput platforms. It uses a relational database to store sequence data and analysis results, which facilitates both the search for specific information and statistical analysis. ESTAP provides for easy viewing of the original and cleansed data, as well as the analysis results via a Web browser. It also allows the data owner to submit selected sequences to dbEST in a semi-automated fashion.