Leanna House

Amphibian skin may select for rare environmental microbes.

Host-microbe symbioses rely on the successful transmission or acquisition of symbionts in each new generation. Amphibians host a diverse cutaneous microbiota, and many of these symbionts appear to be mutualistic and may limit infection by the chytrid fungus, Batrachochytrium dendrobatidis, which has caused global amphibian population declines and extinctions in recent decades. Using bar-coded 454 pyrosequencing of the 16S rRNA gene, we addressed the question of symbiont transmission by examining variation in amphibian skin microbiota across species and sites and in direct relation to environmental microbes. Although acquisition of environmental microbes occurs in some host-symbiont systems, this has not been extensively examined in free-living vertebrate-microbe symbioses. Juvenile bullfrogs (Rana catesbeiana), adult red-spotted newts (Notophthalmus viridescens), pond water and pond substrate were sampled at a single pond to examine host-specificity and potential environmental transmission of microbiota. To assess population level variation in skin microbiota, adult newts from two additional sites were also sampled. Cohabiting bullfrogs and newts had distinct microbial communities, as did newts across the three sites. The microbial communities of amphibians and the environment were distinct; there was very little overlap in the amphibians’ core microbes and the most abundant environmental microbes, and the relative abundances of OTUs that were shared by amphibians and the environment were inversely related. These results suggest that, in a host species-specific manner, amphibian skin may select for microbes that are generally in low abundance in the environment.

Observation-level interaction with statistical models for visual analytics

In visual analytics, sensemaking is facilitated through interactive visual exploration of data. Throughout this dynamic process, users combine their domain knowledge with the dataset to create insight. Therefore, visual analytic tools exist that aid sensemaking by providing various interaction techniques that focus on allowing users to change the visual representation through adjusting parameters of the underlying statistical model. However, we postulate that the process of sensemaking is not focused on a series of parameter adjustments, but instead, a series of perceived connections and patterns within the data. Thus, how can models for visual analytic tools be designed, so that users can express their reasoning on observations (the data), instead of directly on the model or tunable parameters? Observation level (and thus “observation”) in this paper refers to the data points within a visualization. In this paper, we explore two possible observation-level interactions, namely exploratory and expressive, within the context of three statistical methods, Probabilistic Principal Component Analysis (PPCA), Multidimensional Scaling (MDS), and Generative Topographic Mapping (GTM). We discuss the importance of these two types of observation level interactions, in terms of how they occur within the sensemaking process. Further, we present use cases for GTM, MDS, and PPCA, illustrating how observation level interaction can be incorporated into visual analytic tools.

Panamanian frog species host unique skin bacterial communities

Vertebrates, including amphibians, host diverse symbiotic microbes that contribute to host disease resistance. Globally, and especially in montane tropical systems, many amphibian species are threatened by a chytrid fungus, Batrachochytrium dendrobatidis (Bd), that causes a lethal skin disease. Bd therefore may be a strong selective agent on the diversity and function of the microbial communities inhabiting amphibian skin. In Panamá, amphibian population declines and the spread of Bd have been tracked. In 2012, we completed a field survey in Panamá to examine frog skin microbiota in the context of Bd infection. We focused on three frog species and collected two skin swabs per frog from a total of 136 frogs across four sites that varied from west to east in the time since Bd arrival. One swab was used to assess bacterial community structure using 16S rRNA amplicon sequencing and to determine Bd infection status, and one was used to assess metabolite diversity, as the bacterial production of anti-fungal metabolites is an important disease resistance function. The skin microbiota of the three Panamanian frog species differed in OTU (operational taxonomic unit, ~bacterial species) community composition and metabolite profiles, although the pattern was less strong for the metabolites. Comparisons between frog skin bacterial communities from Panamá and the US suggest broad similarities at the phylum level, but key differences at lower taxonomic levels. In our field survey in Panamá, across all four sites, only 35 individuals (~26%) were Bd infected. There was no clustering of OTUs or metabolite profiles based on Bd infection status and no clear pattern of west-east changes in OTUs or metabolite profiles across the four sites. Overall, our field survey data suggest that different bacterial communities might be producing broadly similar sets of metabolites across frog hosts and sites. Community structure and function may not be as tightly coupled in these skin symbiont microbial systems as it is in many macro-systems.

Community Structure and Function of Amphibian Skin Microbes: An Experiment with Bullfrogs Exposed to a Chytrid Fungus

The vertebrate microbiome contributes to disease resistance, but few experiments have examined the link between microbiome community structure and disease resistance functions. Chytridiomycosis, a major cause of amphibian population declines, is a skin disease caused by the fungus, Batrachochytrium dendrobatidis (Bd). In a factorial experiment, bullfrog skin microbiota was reduced with antibiotics, augmented with an anti-Bd bacterial isolate (Janthinobacterium lividum), or unmanipulated, and individuals were then either exposed or not exposed to Bd. We found that the microbial community structure of individual frogs prior to Bd exposure influenced Bd infection intensity one week following exposure, which, in turn, was negatively correlated with proportional growth during the experiment. Microbial community structure and function differed among unmanipulated, antibiotic-treated, and augmented frogs only when frogs were exposed to Bd. Bd is a selective force on microbial community structure and function, and beneficial states of microbial community structure may serve to limit the impacts of infection.

Second-Order Exchangeability Analysis for Multimodel Ensembles

The challenge of understanding complex systems often gives rise to a multiplicity of models. It is natural to consider whether the outputs of these models can be combined to produce a system prediction that is more informative than the output of any one of the models taken in isolation. And, in particular, to consider the relationship between the spread of model outputs and system uncertainty. We describe a statistical framework for such a combination, based on the exchangeability of the models, and their coexchangeability with the system. We demonstrate the simplest implementation of our framework in the context of climate prediction. Throughout we work entirely in means and variances to avoid the necessity of specifying higher-order quantities for which we often lack well-founded judgments.

Accurate characterization of delay discounting: a multiple model approach using approximate Bayesian model selection and a unified discounting measure.

The study of delay discounting, or valuation of future rewards as a function of delay, has contributed to understanding the behavioral economics of addiction. Accurate characterization of discounting can be furthered by statistical model selection given that many functions have been proposed to measure future valuation of rewards. The present study provides a convenient Bayesian model selection algorithm that selects the most probable discounting model among a set of candidate models chosen by the researcher. The approach assigns the most probable model for each individual subject. Importantly, effective delay 50 (ED50) functions as a suitable unifying measure that is computable for and comparable between a number of popular functions, including both one- and two-parameter models. The combined model selection/ED50 approach is illustrated using empirical discounting data collected from a sample of 111 undergraduate students with models proposed by Laibson (1997); Mazur (1987); Myerson & Green (1995); Rachlin (2006); and Samuelson (1937). Computer simulation suggests that the proposed Bayesian model selection approach outperforms the single model approach when data truly arise from multiple models. When a single model underlies all participant data, the simulation suggests that the proposed approach fares no worse than the single model approach.

Bayesian visual analytics: BaVA

Leman et al. and Endert et al. develop an interactive data visualization framework called visual to parametric interaction V2PI. With V2PI, experts may explore data visually assess multiple data visualizations based on their judgments and an underlying data analytic method. Specifically, V2PI offers a deterministic procedure to quantify expert judgments and update analytical parameters to create new data visualizations. In this article, we explain V2PI from a probabilistic perspective and develop Bayesian visual analytics BaVA. We model data probabilistically, develop parallels between quantifying expert judgments and eliciting prior distributions from experts, and justify how we update parameters using Bayesian sequential updating. We apply BaVA using two linear projections methods to assess simulated and real-world datasets.

Multi-model semantic interaction for text analytics

Semantic interaction offers an intuitive communication mechanism between human users and complex statistical models. By shielding the users from manipulating model parameters, they focus instead on directly manipulating the spatialization, thus remaining in their cognitive zone. However, this technique is not inherently scalable past hundreds of text documents. To remedy this, we present the concept of multi-model semantic interaction, where semantic interactions can be used to steer multiple models at multiple levels of data scale, enabling users to tackle larger data problems. We also present an updated visualization pipeline model for generalized multi-model semantic interaction. To demonstrate multi-model semantic interaction, we introduce StarSPIRE, a visual text analytics prototype that transforms user interactions on documents into both small-scale display layout updates as well as large-scale relevancy-based document selection.

Visual to Parametric Interaction (V2PI).

Typical data visualizations result from linear pipelines that start by characterizing data using a model or algorithm to reduce the dimension and summarize structure, and end by displaying the data in a reduced dimensional form. Sensemaking may take place at the end of the pipeline when users have an opportunity to observe, digest, and internalize any information displayed. However, some visualizations mask meaningful data structures when model or algorithm constraints (e.g., parameter specifications) contradict information in the data. Yet, due to the linearity of the pipeline, users do not have a natural means to adjust the displays. In this paper, we present a framework for creating dynamic data displays that rely on both mechanistic data summaries and expert judgement. The key is that we develop both the theory and methods of a new human-data interaction to which we refer as “ Visual to Parametric Interaction” (V2PI). With V2PI, the pipeline becomes bi-directional in that users are embedded in the pipeline; users learn from visualizations and the visualizations adjust to expert judgement. We demonstrate the utility of V2PI and a bi-directional pipeline with two examples.

Bayesian nonparametric models for peak identification in MALDI-TOF mass spectroscopy

We present a novel nonparametric Bayesian approach based on Levy Adaptive Regression Kernels (LARK) to model spectral data arising from MALDI-TOF (Matrix Assisted Laser Desorption Ionization Time-of-Flight) mass spectrometry. This model based approach provides identification and quantification of proteins through model parameters that are directly interpretable as the number of proteins, mass and abundance of proteins and peak resolution, while having the ability to adapt to unknown smoothness as in wavelet based methods. Informative prior distributions on resolution are key to distinguishing true peaks from background noise and resolving broad peaks into individual peaks for multiple protein species. Posterior distributions are obtained using a reversible jump Markov chain Monte Carlo algorithm and provide inference about the number of peaks (proteins), their masses and abundance. We show through simulation studies that the procedure has desirable true-positive and false-discovery rates. Finally, we illustrate the method on five example spectra: a blank spectrum, a spectrum with only the matrix of a low-molecularweight substance used to embed target proteins, a spectrum with known proteins, and a single spectrum and average of ten spectra from an individual lung cancer patient.