Quantifying conditional probability tables in Bayesian networks: Bayesian regression for scenario-based encoding of elicited expert assessments on feral pig habitat.

Bayesian networks are now widespread for modelling uncertain knowledge. They graph probabilistic relationships, which are quantified using conditional probability tables (CPTs). When empirical data are unavailable, experts may specify CPTs. Here we propose novel methodology for quantifying CPTs: a Bayesian statistical approach to both elicitation and encoding of expert-specified probabilities, in a way that acknowledges their uncertainty. We illustrate this new approach using a case study describing habitat most at risk from feral pigs. For complicated CPTs, it is difficult to elicit all scenarios (CPT entries). Like the CPT Calculator software program, we ask about a few scenarios (e.g. under a one-factor-at-a-time design) to reduce the experts' workload. Unlike CPT Calculator, we adopt a global rather than local regression to 'fill out' CPT entries. Unlike other methods for scenario-based elicitation for regression, we capture uncertainty about each probability in a sequence that explicitly controls biases and enhances interpretation. Furthermore, to utilize all elicited information, we introduce Bayesian rather than Classical generalised linear modelling (GLM). For large CPTs (e.g. >3 levels per parent) we show Bayesian GLM supports richer inference, particularly on interactions, even with few scenarios, providing more information regarding accuracy of encoding.

Risk Associated with the Release of Wolbachia-Infected Aedes aegypti Mosquitoes into the Environment in an Effort to Control Dengue.

BACKGROUND: In an effort to eliminate dengue, a successful technology was developed with the stable introduction of the obligate intracellular bacteria Wolbachia pipientis into the mosquito Aedes aegypti to reduce its ability to transmit dengue fever due to life shortening and inhibition of viral replication effects. An analysis of risk was required before considering release of the modified mosquito into the environment. METHODS: Expert knowledge and a risk assessment framework were used to identify risk associated with the release of the modified mosquito. Individual and group expert elicitation was performed to identify potential hazards. A Bayesian network (BN) was developed to capture the relationship between hazards and the likelihood of events occurring. Risk was calculated from the expert likelihood estimates populating the BN and the consequence estimates elicited from experts. RESULTS: The risk model for "Don't Achieve Release" provided an estimated 46% likelihood that the release would not occur by a nominated time but generated an overall risk rating of very low. The ability to obtain compliance had the greatest influence on the likelihood of release occurring. The risk model for "Cause More Harm" provided a 12.5% likelihood that more harm would result from the release, but the overall risk was considered negligible. The efficacy of mosquito management had the most influence, with the perception that the threat of dengue fever had been eliminated, resulting in less household mosquito control, and was scored as the highest ranked individual hazard (albeit low risk). CONCLUSIONS: The risk analysis was designed to incorporate the interacting complexity of hazards that may affect the release of the technology into the environment. The risk analysis was a small, but important, implementation phase in the success of this innovative research introducing a new technology to combat dengue transmission in the environment.

Intrinsic Antimicrobial Resistance Determinants in the Superbug Pseudomonas aeruginosa.

UNLABELLED: Antimicrobial-resistant bacteria pose a serious threat in the clinic. This is particularly true for opportunistic pathogens that possess high intrinsic resistance. Though many studies have focused on understanding the acquisition of bacterial resistance upon exposure to antimicrobials, the mechanisms controlling intrinsic resistance are not well understood. In this study, we subjected the model opportunistic superbug Pseudomonas aeruginosa to 14 antimicrobials under highly controlled conditions and assessed its response using expression- and fitness-based genomic approaches. Our results reveal that gene expression changes and mutant fitness in response to sub-MIC antimicrobials do not correlate on a genomewide scale, indicating that gene expression is not a good predictor of fitness determinants. In general, fewer fitness determinants were identified for antiseptics and disinfectants than for antibiotics. Analysis of gene expression and fitness data together allowed the prediction of antagonistic interactions between antimicrobials and insight into the molecular mechanisms controlling these interactions. IMPORTANCE: Infections involving multidrug-resistant pathogens are difficult to treat because the therapeutic options are limited. These infections impose a significant financial burden on infected patients and on health care systems. Despite years of antimicrobial resistance research, we lack a comprehensive understanding of the intrinsic mechanisms controlling antimicrobial resistance. This work uses two fine-scale genomic approaches to identify genetic loci important for antimicrobial resistance of the opportunistic pathogen Pseudomonas aeruginosa. Our results reveal that antibiotics have more resistance determinants than antiseptics/disinfectants and that gene expression upon exposure to antimicrobials is not a good predictor of these resistance determinants. In addition, we show that when used together, genomewide gene expression and fitness profiling can provide mechanistic insights into multidrug resistance mechanisms.

Essential genome of Pseudomonas aeruginosa in cystic fibrosis sputum.

Defining the essential genome of bacterial pathogens is central to developing an understanding of the biological processes controlling disease. This has proven elusive for Pseudomonas aeruginosa during chronic infection of the cystic fibrosis (CF) lung. In this paper, using a Monte Carlo simulation-based method to analyze high-throughput transposon sequencing data, we establish the P. aeruginosa essential genome with statistical precision in laboratory media and CF sputum. Reconstruction of the global requirements for growth in CF sputum compared with defined growth conditions shows that the latter requires several cofactors including biotin, riboflavin, and pantothenate. Comparison of P. aeruginosa strains PAO1 and PA14 demonstrates that essential genes are primarily restricted to the core genome; however, some orthologous genes in these strains exhibit differential essentiality. These results indicate that genes with similar molecular functions may have distinct genetic roles in different P. aeruginosa strains during growth in CF sputum. We also show that growth in a defined growth medium developed to mimic CF sputum yielded virtually identical fitness requirements to CF sputum, providing support for this medium as a relevant in vitro model for CF microbiology studies.

Mechanisms of synergy in polymicrobial infections.

Communities of microbes can live almost anywhere and contain many different species. Interactions between members of these communities often determine the state of the habitat in which they live. When these habitats include sites on the human body, these interactions can affect health and disease. Polymicrobial synergy can occur during infection, in which the combined effect of two or more microbes on disease is worse than seen with any of the individuals alone. Powerful genomic methods are increasingly used to study microbial communities, including metagenomics to reveal the members and genetic content of a community and metatranscriptomics to describe the activities of community members. Recent efforts focused toward a mechanistic understanding of these interactions have led to a better appreciation of the precise bases of polymicrobial synergy in communities containing bacteria, eukaryotic microbes, and/or viruses. These studies have benefited from advances in the development of in vivo models of polymicrobial infection and modern techniques to profile the spatial and chemical bases of intermicrobial communication. This review describes the breadth of mechanisms microbes use to interact in ways that impact pathogenesis and techniques to study polymicrobial communities.