We present a generative model for simultaneously clustering documents and terms. Our model is a four-level hierarchical Bayesian model, in which each document is modeled as a random mixture of document topics , where each topic is a distribution over some segments of the text. Each of these segments in the document can be modeled as a mixture of word topics where each topic is a distribution over words. We present efficient approximate inference techniques based on Markov Chain Monte Carlo method and a Moment-Matching algorithm for empirical Bayes parameter estimation. We report results in document modeling, document and term clustering, comparing to other topic models,

Community Question Answering (CQA) websites provide a rapidly growing source of information in many areas. This rapid growth, while offering new opportunities, puts forward new challenges. In most CQA implementations there is little effort in directing new questions to the right group of experts. This means that experts are not provided with questions matching their expertise, and therefore new matching questions may be missed and not receive a proper answer. We focus on finding experts for a newly posted question. We investigate the suitability of two statistical topic models for solving this issue and compare these methods against more traditional Information Retrieval approaches. We show that for a dataset constructed from the Stackoverflow website, these topic models outperform other methods in retrieving a candidate set of best experts for a question. We also show that the Segmented Topic Model gives consistently better performance compared to the Latent Dirichlet Allocation Model.

BackgroundMicrobiome samples often represent mixtures of communities, where each community is composed of overlapping assemblages of species. Such mixtures are complex, the number of species is huge and abundance information for many species is often sparse. Classical methods have a limited value for identifying complex features within such data.ResultsHere, we describe a novel hierarchical model for Bayesian inference of microbial communities (BioMiCo). The model takes abundance data derived from environmental DNA, and models the composition of each sample by a two-level hierarchy of mixture distributions constrained by Dirichlet priors. BioMiCo is supervised, using known features for samples and appropriate prior constraints to overcome the challenges posed by many variables, sparse data, and large numbers of rare species. The model is trained on a portion of the data, where it learns how assemblages of species are mixed to form communities and how assemblages are related to the known features of each sample. Training yields a model that can predict the features of new samples. We used BioMiCo to build models for three serially sampled datasets and tested their predictive accuracy across different time points. The first model was trained to predict both body site (hand, mouth, and gut) and individual human host. It was able to reliably distinguish these features across different time points. The second was trained on vaginal microbiomes to predict both the Nugent score and individual human host. We found that women having normal and elevated Nugent scores had distinct microbiome structures that persisted over time, with additional structure within women having elevated scores. The third was trained for the purpose of assessing seasonal transitions in a coastal bacterial community. Application of this model to a high-resolution time series permitted us to track the rate and time of community succession and accurately predict known ecosystem-level events.ConclusionBioMiCo provides a framework for learning the structure of microbial communities and for making predictions based on microbial assemblages. By training on carefully chosen features (abiotic or biotic), BioMiCo can be used to understand and predict transitions between complex communities composed of hundreds of microbial species.Electronic supplementary materialThe online version of this article (doi:10.1186/s40168-015-0073-x) contains supplementary material, which is available to authorized users.

Wetting experiments show pure graphene to be weakly hydrophilic, but its contact angle (CA) also reflects the character of the supporting material. Measurements and molecular dynamics simulations on suspended and supported graphene often reveal a CA reduction due to the presence of the supporting substrate. A similar reduction is consistently observed when graphene is wetted from both sides. The effect has been attributed to transparency to molecular interactions across the graphene sheet; however, the possibility of substrate-induced graphene polarization has also been considered. Computer simulations of CA on graphene have so far been determined by ignoring the material's conducting properties. We improve the graphene model by incorporating its conductivity according to the constant applied potential molecular dynamics. Using this method, we compare the wettabilities of suspended graphene and graphene supported by water by measuring the CA of cylindrical water drops on the sheets. The inclusion of graphene conductivity and concomitant polarization effects leads to a lower CA on suspended graphene, but the CA reduction is significantly bigger when the sheets are also wetted from the opposite side. The stronger adhesion is accompanied by a profound change in the correlations among water molecules across the sheet. While partial charges on water molecules interacting across an insulator sheet attract charges of the opposite sign, apparent attraction among like charges is manifested across the conducting graphene. The change is associated with graphene polarization, as the image charges inside the conductor attract equally signed partial charges of water molecules on both sides of the sheet. Additionally, using a nonpolar liquid (diiodomethane), we affirm a detectable wetting translucency when liquid−liquid forces are dominated by dispersive interactions. Our findings are important for predictive modeling toward a variety of applications including sensors, fuel cell membranes, water filtration, and graphene-based electrode materials in high-performance supercapacitors.

Metagenomics yields enormous numbers of microbial sequences that can be assigned a metabolic function. Using such data to infer community-level metabolic divergence is hindered by the lack of a suitable statistical framework. Here, we describe a novel hierarchical Bayesian model, called BiomeNet (Bayesian inference of metabolic networks), for inferring differential prevalence of metabolic subnetworks among microbial communities. To infer the structure of community-level metabolic interactions, BiomeNet applies a mixed-membership modelling framework to enzyme abundance information. The basic idea is that the mixture components of the model (metabolic reactions, subnetworks, and networks) are shared across all groups (microbiome samples), but the mixture proportions vary from group to group. Through this framework, the model can capture nested structures within the data. BiomeNet is unique in modeling each metagenome sample as a mixture of complex metabolic systems (metabosystems). The metabosystems are composed of mixtures of tightly connected metabolic subnetworks. BiomeNet differs from other unsupervised methods by allowing researchers to discriminate groups of samples through the metabolic patterns it discovers in the data, and by providing a framework for interpreting them. We describe a collapsed Gibbs sampler for inference of the mixture weights under BiomeNet, and we use simulation to validate the inference algorithm. Application of BiomeNet to human gut metagenomes revealed a metabosystem with greater prevalence among inflammatory bowel disease (IBD) patients. Based on the discriminatory subnetworks for this metabosystem, we inferred that the community is likely to be closely associated with the human gut epithelium, resistant to dietary interventions, and interfere with human uptake of an antioxidant connected to IBD. Because this metabosystem has a greater capacity to exploit host-associated glycans, we speculate that IBD-associated communities might arise from opportunist growth of bacteria that can circumvent the host's nutrient-based mechanism for bacterial partner selection.