MOTIVATION: Next-generation sequencing coupled with metagenomics has led to the rapid growth of sequence databases and enabled a new branch of microbiology called comparative metagenomics. Comparative metagenomic analysis studies compositional patterns within and between different environments providing a deep insight into the structure and function of complex microbial communities. It is a fast growing field that requires the development of novel supervised learning techniques for addressing challenges associated with metagenomic data, e.g. sensitivity to the choice of sequence similarity cutoff used to define operational taxonomic units (OTUs), high dimensionality and sparsity of the data and so forth. On the other hand, the natural properties of microbial community data may provide useful information about the structure of the data. For example, similarity between species encoded by a phylogenetic tree captures the relationship between OTUs and may be useful for the analysis of complex microbial datasets where the diversity patterns comprise features at multiple taxonomic levels. Even though some of the challenges have been addressed by learning algorithms in the literature, none of the available methods take advantage of the inherent properties of metagenomic data. RESULTS: We proposed a novel supervised classification method for microbial community samples, where each sample is represented as a set of OTU frequencies, which takes advantage of the natural structure in microbial community data encoded by a phylogenetic tree. This model allows us to take advantage of environment-specific compositional patterns that may contain features at multiple granularity levels. Our method is based on the multinomial logistic regression model with a tree-guided penalty function. Additionally, we proposed a new simulation framework for generating 16S ribosomal RNA gene read counts that may be useful in comparative metagenomics research. Our experimental results on simulated and real data show that the phylogenetic information used in our method improves the classification accuracy. AVAILABILITY AND IMPLEMENTATION: http://www.cs.ucr.edu/~tanaseio/metaphyl.htm.

BACKGROUND: PRISE2 is a new software tool for designing sequence-selective PCR primers and probes. To achieve high level of selectivity, PRISE2 allows the user to specify a collection of target sequences that the primers are supposed to amplify, as well as non-target sequences that should not be amplified. The program emphasizes primer selectivity on the 3' end, which is crucial for selective amplification of conserved sequences such as rRNA genes. In PRISE2, users can specify desired properties of primers, including length, GC content, and others. They can interactively manipulate the list of candidate primers, to choose primer pairs that are best suited for their needs. A similar process is used to add probes to selected primer pairs. More advanced features include, for example, the capability to define a custom mismatch penalty function. PRISE2 is equipped with a graphical, user-friendly interface, and it runs on Windows, Macintosh or Linux machines. RESULTS: PRISE2 has been tested on two very similar strains of the fungus Dactylella oviparasitica, and it was able to create highly selective primers and probes for each of them, demonstrating the ability to create useful sequence-selective assays. CONCLUSIONS: PRISE2 is a user-friendly, interactive software package that can be used to design high-quality selective primers for PCR experiments. In addition to choosing primers, users have an option to add a probe to any selected primer pair, enabling design of Taqman and other primer-probe based assays. PRISE2 can also be used to design probes for FISH and other hybridization-based assays.

BACKGROUND: Population levels of microbial phylotypes can be examined using a hybridization-based method that utilizes a small set of computationally-designed DNA probes targeted to a gene common to all. Our previous algorithm attempts to select a set of probes such that each training sequence manifests a unique theoretical hybridization pattern (a binary fingerprint) to a probe set. It does so without taking into account similarity between training gene sequences or their putative taxonomic classifications, however. We present an improved algorithm for probe set selection that utilizes the available taxonomic information of training gene sequences and attempts to choose probes such that the resultant binary fingerprints cluster into real taxonomic groups. RESULTS: Gene sequences manifesting identical fingerprints with probes chosen by the new algorithm are more likely to be from the same taxonomic group than probes chosen by the previous algorithm. In cases where they are from different taxonomic groups, underlying DNA sequences of identical fingerprints are more similar to each other in probe sets made with the new versus the previous algorithm. Complete removal of large taxonomic groups from training data does not greatly decrease the ability of probe sets to distinguish those groups. CONCLUSIONS: Probe sets made from the new algorithm create fingerprints that more reliably cluster into biologically meaningful groups. The method can readily distinguish microbial phylotypes that were excluded from the training sequences, suggesting novel microbes can also be detected.

Polony technology is a low-cost, high-throughput platform employed in several applications such as DNA sequencing, haplotyping and alternative pre-mRNA splicing analysis. Owing to their random placement, however, overlapping polonies occur often and may result in inaccurate or unusable data. Accurately identifying polony positions and sizes is essential for maximizing the quantity and quality of data aquired in an image, however, most existing identification algorithms do not handle overlapping polonies well. In this paper, we present a novel polony identification approach combining both a Gaussian Mixture Model (GMM) and the Expectation-Maximization (EM) algorithm. Experiments on simulated and real images of highly overlapping polonies show that our algorithm has a 10% to 20% increase in recall compared with the existing algorithms, while keeping precision at the same level.

MOTIVATION: Oligonucleotide fingerprinting of ribosomal RNA genes (OFRG) is a procedure that sorts rRNA gene (rDNA) clones into taxonomic groups through a series of hybridization experiments. The hybridization signals are classified into three discrete values 0, 1 and N, where 0 and 1, respectively, specify negative and positive hybridization events and N designates an uncertain assignment. This study examined various approaches for classifying the values including Bayesian classification with normally distributed signal data, Bayesian classification with the exponentially distributed data, and with gamma distributed data, along with tree-based classification. All classification data were clustered using the unweighted pair group method with arithmetic mean. RESULTS: The performance of each classification/clustering procedure was compared with results from known reference data. Comparisons indicated that the approach using the Bayesian classification with normal densities followed by tree clustering out-performed all others. The paper includes a discussion of how this Bayesian approach may be useful for the analysis of gene expression data.