Clustering consistency with Dirichlet process mixtures

Dirichlet process mixtures are flexible nonparametric models, particularly suited to density estimation and probabilistic clustering. In this work we study the posterior distribution induced by Dirichlet process mixtures as the sample size increases, and more specifically focus on consistency for the unknown number of clusters when the observed data are generated from a finite mixture. Crucially, we consider the situation where a prior is placed on the concentration parameter of the underlying Dirichlet process. Previous findings in the literature suggest that Dirichlet process mixtures are typically not consistent for the number of clusters if the concentration parameter is held fixed and data come from a finite mixture. Here we show that consistency for the number of clusters can be achieved if the concentration parameter is adapted in a fully Bayesian way, as commonly done in practice. Our results are derived for data coming from a class of finite mixtures, with mild assumptions on the prior for the concentration parameter and for a variety of choices of likelihood kernels for the mixture.

Pulsed Electric Field Processing of Red Grapes (cv. Rondinella): Modifications of Phenolic Fraction and Effects on Wine Evolution

Pulsed electric field (PEF) is a non-thermal technology able to promote color and polyphenols extraction from grape skins. Most of the publications about PEF in winemaking report data concerning international varieties, poorly considering minor cultivars and the medium/long-term effects of the treatment on wine composition during storage. PEF was applied at different specific energies (2, 10, and 20 kJ kg-1) on grapes of the low-color red cv. Rondinella, after crushing-destemming. Pressing yield, the evolution of color, and total phenolic index (TPI) were measured during skin maceration. Moreover, the wines were characterized for basic compositional parameters, color, anthocyanin profile, phenolic composition (glories indices), metal content (Fe, Cr, and Ni), and sensory characters, two and twelve months after the processing, in comparison with untreated samples and pectolytic enzymes (PE). PEF did not affect fermentation evolution, nor did it modify wine basic composition or metal content. Treatments at 10 and 20 kJ kg-1 led to higher color and TPI in wines, in comparison to PE, because of increased content of anthocyanins and tannins. The sensory evaluation confirmed these findings. Modifications remained stable in wines after twelve months. Glories indices and vitisin A content highlighted greater potential stability of wine color in PEF-treated wines.

Scalable inference for crossed random effects models

We develop methodology and complexity theory for Markov chain Monte Carlo algorithms used in inference for crossed random effects models in modern analysis of variance. We consider a plain Gibbs sampler and propose a simple modification, referred to as a collapsed Gibbs sampler. Under some balancedness conditions on the data designs and assuming that precision hyperparameters are known, we demonstrate that the plain Gibbs sampler is not scalable, in the sense that its complexity is worse than proportional to the number of parameters and data, but the collapsed Gibbs sampler is scalable. In simulated and real datasets we show that the explicit convergence rates predicted by our theory closely match the computable, but nonexplicit rates in cases where the design assumptions are violated. We also show empirically that the collapsed Gibbs sampler extended to sample precision hyperparameters significantly outperforms alternative state-of-the-art algorithms.

Unbiased approximations of products of expectations

We consider the problem of approximating the product of $n$ expectations with respect to a common probability distribution $\mu$. Such products routinely arise in statistics as values of the likelihood in latent variable models. Motivated by pseudo-marginal Markov chain Monte Carlo schemes, we focus on unbiased estimators of such products. The standard approach is to sample $N$ particles from $\mu$ and assign each particle to one of the expectations; this is wasteful and typically requires the number of particles to grow quadratically with the number of expectations. We propose an alternative estimator that approximates each expectation using most of the particles while preserving unbiasedness, which is computationally more efficient when the cost of simulations greatly exceeds the cost of likelihood evaluations. We carefully study the properties of our proposed estimator, showing that in latent variable contexts it needs only ${O} (n)$ particles to match the performance of the standard approach with ${O}(n^{2})$ particles. We demonstrate the procedure on two latent variable examples from approximate Bayesian computation and single-cell gene expression analysis, observing computational gains by factors of about 25 and 450, respectively.

Bovine anaplasmosis and tick-borne pathogens in cattle of the Galapagos Islands

A cross-sectional study was conducted to determine the species of Anaplasma spp. and estimate its prevalence in cattle of the three main cattle-producing Galapagos Islands (Santa Cruz, San Cristóbal and Isabela) using indirect PCR assays, genetic sequencing and ELISA. Ticks were also collected from cattle and scanned for 47 tick-borne pathogens in a 48 × 48 real-time PCR chip. A mixed effects logistic regression was performed to identify potential risk factors explaining Anaplasma infection in cattle. A. phagocytophilum was not detected in any of the tested animals. Genetic sequencing allowed detection of A. platys-like strains in 11 (36.7%) of the 30 Anaplasma spp.-positive samples analysed. A. marginale was widespread in the three islands with a global between-herd prevalence of 100% [89; 100]_{95% CI} and a median within-herd prevalence of 93%. A significant association was found between A. marginale infection and age with higher odds of being positive for adults (OR = 3.3 [1.2; 9.9]_{95% Bootstrap CI} ). All collected ticks were identified as Rhipicephalus microplus. A. marginale, Babesia bigemina, Borrelia theileri and Francisella-like endosymbiont were detected in tick pools. These results show that the Galapagos Islands are endemic for A. marginale.

Estimation of French cattle herd immunity against bluetongue serotype 8 at the time of its re-emergence in 2015

BACKGROUND

From 2006 to 2010, France experienced two bluetongue epidemics caused by serotype 1 (BTV-1) and 8 (BTV-8) which were controlled by mass vaccination campaigns. After five years without any detected cases, a sick ram was confirmed in August 2015 to be infected by a BTV-8 strain almost identical to that circulating during the previous outbreak. By then, part of the French cattle population was expected to be still protected, since bluetongue antibodies are known to last for many years after natural infection or vaccination. The objective of this study was to estimate the proportion of cattle in France still immune to BTV-8 at the time of its re-emergence in 2015.

RESULTS

We used BTV group-specific cELISA results from 8525 cattle born before the vaccination ban in 2013 and 15,799 cattle born after the ban. Samples were collected from January to April 2016 to estimate seroprevalence per birth cohort. The overall seroprevalence in cattle at national and local levels was extrapolated from seroprevalence results per birth cohort and their respective proportion at each level. To indirectly assess pre-immune status of birth cohorts, we computed prevalence per birth cohort on infected farms in autumn 2015 using 1377 RT-PCR results. These revealed limited BTV circulation in 2015. Seroprevalence per birth cohort was likely to be connected to past exposure to natural infection and/or vaccination with higher seroprevalence levels in older animals. A seroprevalence of 95% was observed for animals born before 2008, of which > 90% were exposed to two compulsory vaccination campaigns in 2008-2010. None of the animals born before 2008 were found to be infected, unlike 19% of the young cattle which had never been vaccinated. This suggests that most ELISA-positive animals were pre-immune to BTV-8. We estimated that 18% (from 12% to 32% per département) of the French cattle population was probably pre-immune in 2015.

CONCLUSIONS

These results strongly suggest a persistence of antibodies for at least 5-6 years after natural infection or vaccination. The herd immunity of the French cattle population probably limited BTV circulation up to 2015, by which time more than 80% of cattle were naive.

Serological status for BTV-8 in French cattle prior to the 2015 re-emergence

Undetected in Europe since 2010, bluetongue virus serotype 8 (BTV-8) re-emerged in August 2015 in Central France. To gain insight into the re-emergence on the French territory, we estimated the seroprevalence in cattle before the detection of BTV-8 in 2015, in areas differentially affected by the current outbreak. A retrospective survey based on the analysis of stored sera was thus conducted in the winter preceding the re-emergence in seven French departments including the one where the virus was first detected. A total of 10,066 sera were retrieved from animals sampled in 444 different herds in winter 2014/15. Between-herd seroprevalence revealed the presence of seropositive animals in almost all herds sampled (97.4%). The animal-level seroprevalence averaged at 44%, with a strong age pattern reflecting the cumulative exposure to both natural infection and to vaccination. A multivariable analysis allowed separating the respective effects of both exposures. A higher proportion of seropositivity risk was attributed to vaccination (67.4%) than to exposure to natural infection (24.2%). The evolution of seroprevalence induced by the two main risk factors in 74 mainland departments was reconstructed between the vaccination ban (2013) and the re-emergence (2015). We showed a striking decrease in seroprevalence with time after the vaccination ban, due to population renewal, which could have facilitated virus transmission leading to the current outbreak situation.

Circulation of bluetongue virus 8 in French cattle, before and after the re-emergence in 2015

Bluetongue virus serotype 8 (BTV-8) re-emerged in Central France in August 2015. The viral strain identified is nearly identical to the one that circulated during the 2006/2009 massive outbreak throughout Europe. To address the question of an undetected BTV-8 circulation on the French territory, a serological study was conducted on young cattle along a transect of seven departments, three of them located in areas where the virus presence had been confirmed by RT-PCR by winter 2015/2016. Sera from 2,565 animals were collected during the winters preceding and following the re-emergence, with 414 animals being sampled in each of the two consecutive years. All samples were tested by competitive ELISA (IDVet) and, when enough serum was available, ELISA-positive samples were confirmed by seroneutralization tests. In areas with infected holdings, seropositive animals were found before the re-emergence (N = 14 of 511), significantly more on the following year (N = 17 of 257), and eight animals (N = 158) seroconverted over 2015. Seropositive animals were also detected as early as winter 2014/2015 in one department without known infected holdings (N = 12 of 150), and in winter 2015/2016 in three of them (N = 21 of 555), where seven animals (N = 154) seroconverted over 2015. These results suggest that BTV-8 may have spread at low levels before the re-emergence, even in areas considered virus-free. Unfortunately, whole blood from the seropositive animals was not available to definitely confirm the virus presence by RT-PCR.

Likely Introduction Date of Schmallenberg Virus into France According to Monthly Serological Surveys in Cattle

To estimate the date of introduction of Schmallenberg virus (SBV) into France, the prevalence of antibodies against the virus was determined monthly in cattle from two northern departments from August 2011 to April 2012. Seropositive cattle were detected from October 2011 in both departments with a prevalence of 55.6% in the westernmost department (Meurthe-et-Moselle) and of 12.7% in the easternmost department (Manche). Schmallenberg virus seroprevalence then increased rapidly to high levels.

Modelling transmission of bovine tuberculosis in red deer and wild boar in Normandy, France

In early 2001, Mycobacterium bovis infection was confirmed in red deer (RD) (Cervus elaphus) shot in Normandy region, France. An epidemiological survey conducted during the following hunting season in two connected forests confirmed the occurrence of the disease in both free-ranging RD and wild boar (WB) (Sus scrofa). This was the first detected bovine tuberculosis outbreak in wildlife in France. We present a simple deterministic age-structured model of the within- and between-species M. bovis transmission in RD and WB populations that distinguishes direct transmission (horizontal and pseudo-vertical) and indirect transmission through contaminated offal left behind by hunters. Results issued from the epidemiological surveys conducted in Normandy forests were used to estimate transmission parameters. Because data for RD and WB populations were not available, population sizes at demographic equilibrium were estimated and used to run the model. We qualitatively tested different control measure scenarios with our model, considering different mortality rates and offal harvesting, to determine which ones affect the success of infection control. The most realistic control scenario would combine the total depopulation of RD and good compliance with offal harvesting, because the model suggests that infected offal left by hunters represents the main transmission source of M. bovis in the field.

A one-year follow-up of antibody response in cattle and sheep after vaccination with serotype 8- and serotype 1-inactivated BT vaccines

Sixteen sheep and 18 cattle were followed up during 1 year to estimate the duration of immunity induced by inactivated bluetongue virus serotype 8 (BTV-8) vaccines (sheep and cattle) and a bluetongue virus serotype 1 (BTV-1) vaccine (cattle) under field conditions using cELISA and seroneutralization test (SNT). Four sheep never seroconverted. Those that seroconverted were all seronegative by BTV-8 SNT at the date of last sampling [378 days post-vaccination (dpv)]. Eight sheep were still positive by competitive ELISA (cELISA) 378 dpv. All the cattle seroconverted. At the end of the study, eight and 11 cattle were still positive by BTV-8 SNT and cELISA, respectively (335 dpv); and nine were still positive by BTV-1 SNT (301 dpv).

Clinical pattern characterization of cattle naturally infected by BTV-8

Forty-one cattle from seven Belgian farms and two French farms confirmed as infected with bluetongue virus serotype 8 (BTV-8) were monitored from the onset of clinical signs to describe the disease pattern and estimate the duration of blood RT-qPCR and competitiveELISA positivity under field conditions. On each visit, blood samples were taken, and a standardized clinical form was filled in for each animal. A clinical score was calculated for every week until the end of clinical signs. A classification and regression tree (CART) analysis was conducted to determine the most important clinical signs every week for the first 7 weeks. The highest scores were recorded within 2 weeks of clinical onset. The first recorded clinical signs were quite obviously visible (lethargy, conjunctivitis, lesions of nasal mucosa, nasal discharge). Skin lesions, a drop in milk production and weight loss appeared later in the course of the disease. A biphasic pattern regarding nasal lesions was noticed: the first peak concerned mainly congestive and ulcerative lesions, whereas the second peak mainly concerned crusty lesions. The median time estimated by survival analysis to obtain negative RT-qPCR results from the onset of clinical signs was 195 days (range 166-213 days) in the 23 cattle included in the analysis. Serological results remained strongly positive until the end of the study. These results should ensure more accurate detection of an emerging infectious disease and are of prime importance in improving the modelling of BTV-8 persistence in Europe.

Bluetongue virus serotype 8: abortion and transplacental transmission in cattle in the Burgundy region, France, 2008-2009

During the incursion of bluetongue virus (BTV) serotype 8 in France in 2007, an increase in the number of abortions in cattle was observed, but the cause was not clearly established. A survey of all the reported cases of abortion in cattle from November 2008 to April 2009 was conducted in the Nièvre district (Burgundy region) to determine the percentage of abortions as a result of BTV-8 and to study factors that could have played a role in BTV-8 transplacental transmission. BTV-8 was present in 16% of the fetuses or newborn calves that died within 48 h, from 780 dams. Dams inseminated before the BTV epizootic peak recorded from July to September 2008 were more likely to have BTV-positive abortions (OR=5.7, P<0.001) and those vaccinated in May or June 2008 were less likely to have BTV-positive abortions (OR=0.3, P=0.01 and OR=0.4, P=0.001, respectively). The gestational month was not a predictor of BTV abortion. In blood or spleen, fetuses/calves from RT-PCR-positive dams had significantly higher RNA concentrations than fetuses/calves from RT-PCR-negative dams. Of the 128 dams that had BTV-positive fetuses or calves, 60% were RT-PCR-negative. BTV-8-positive fetuses/calves were significantly more frequent (n=42 vs n=21, P=0.082) amongst those showing clinical signs or lesions suggestive of cerebral damage.

Patterns of lesions of bovine tuberculosis in wild red deer and wild boar

The data obtained from a survey of Mycobacterium bovis infection in wild red deer (Cervus elaphus) and wild boar (Sus scrofa) conducted in France in the 2005/06 hunting season were used to describe and quantify the pathological findings in the two species. The red deer had caseous abscessed lesions in their organs and lymph nodes, whereas in the wild boar the lesions were predominantly caseocalcareous and occurred mainly in the lymph nodes. The severity of the gross tuberculosis-like lesions was estimated on the basis of a numerical score. The significant difference between the distribution of the scores in the two species indicated that the disease was more serious in the red deer than in the wild boar. Unlike the red deer, the wild boar did not show a generalised pattern of disease. Among the lymph nodes examined systematically, gross lesions were most frequently observed in the mesenteric lymph nodes in the red deer and in the retropharyngeal lymph nodes in the wild boar. In both species, the presence of gross lesions showed the closest agreement with the isolation of M bovis from the same lymph nodes. The different patterns of the lesions of tuberculosis in the two species suggest that red deer might play an important role in the intraspecies and interspecies dissemination of the infection, whereas in wild boar the spread of the infection would be more likely to be restricted to other wild boar.

Expression of doxorubicin-daunorubicin resistance genes in different anthracycline-producing mutants of Streptomyces peucetius

Two DNA fragments, ric1 and ric2, were isolated from the Streptomyces peucetius 7600 mutant, which produces daunorubicin and doxorubicin, on the basis of their abilities to confer doxorubicin and daunorubicin resistance to Streptomyces lividans. These two fragments are unrelated by restriction mapping and do not show any homology by Southern analysis, yet both of them increase the level of resistance 10-fold in transformed S. lividans. Functional analysis revealed that ric1 also contains two genes of daunorubicin biosynthesis: one coding for the aklavinone C-11 hydroxylase and the other corresponding to the putative dnrR2 regulatory gene of wild-type S. peucetius ATCC 29050 (K. J. Stutzman-Engwall, S. L. Otten, and C. R. Hutchinson, J. Bacteriol. 174:144-154, 1992). Northern (RNA) blot experiments, performed with a ric1 fragment containing daunorubicin-doxorubicin resistance gene(s), revealed a transcript of about 2,100 nucleotides that is present only during the phase of anthracycline metabolite production. The amount of this transcript is higher in strain 7600 than in strain 7900, a mutant which produces 5-fold more daunorubicin and 10-fold less doxorubicin than 7600. Furthermore, two 7900-derived blocked mutants, 8600 and 9700, do not express the 2,100-nucleotide transcript in spite of the absence of gross rearrangements in the ric1 region such as occur with the 7900 parental strain.

On-line motion analysis using a TV system

An electronic method, using a TV system, which analyzes on-line the motion of various things, is described. The single sampled images may be digitized and stored in a minicomputer memory at TV rates, but this limited sampling frequency may be increased with the aid of a triggered stroboflash. The analysis of the acquired data may be made after the acquisition of either an image sequence or a single frame, according to program time and frame sampling frequency.

Picture coding possibility without point by point sampling

It is emphasized that, for many applications in the field of automatic image analysis, it is not useful to sample them in a point by point ordered array. In fact a run-coding may be executed during the scanning process itself of the images. Examples are reported to show the efficacy of the method.