Online Bayesian Analysis with BEAST 2

Bouckaert, Remco R.; Collienne, Lena; Gavryushkin, Alex

doi:10.1101/2022.05.03.490538

Cited by 13 publications

(19 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The more samples, the more accurate the matrix estimate but the longer it takes for the algorithm to finish. In that sense, it nicely fits the class of online algorithms (Bouckaert et al, 2022) which is what we would be doing with full MCMC. So, even though a small MCMC sample will not be sufficient to accurately represent the distribution of κ and statistics like 95% HPDs, there will be sufficient information to accurately estimate two parameters of a log-normal, from which statistics like 95% HPDs then can accurately be inferred.…”

Section: Discussionmentioning

confidence: 85%

“…The more samples, the more accurate the matrix estimate but the longer it takes for the algorithm to finish. In that sense, it nicely fits the class of online algorithms (Bouckaert et al, 2022) that at any time can provide an answer, but the answer becomes more accurate the longer the algorithm runs. Detecting when longer runs do not increase accuracy of the posterior any more is an open question where automated convergence techniques employed for standard MCMC (Berling et al, 2023) could be exploited.…”

Section: /17mentioning

confidence: 79%

See 1 more Smart Citation

Variational Bayesian Phylogenies through Matrix Representation of Tree Space

Bouckaert

2023

Preprint

View full text Add to dashboard Cite

Variational Bayesian methods work well when the distribution of interest can be represented by a Euclidian space using a bijective transformation between phylogenies and their representation. Until now, full phylogenetic inference has been hampered by the lack of such transformation. Here, we study the distance matrix as a representation of a phylogeny through single link clustering. Though such representation does not map to unique trees, restriction to a subspace we call a "cube", makes the representation bijective at the cost of not being able to represent all possible trees. We introduce a variational Bayesian algorithm "cubeVB" based on MCMC and show through well calibrated simulation study that it is possible to recover parameters of interest like tree height and length. Although a cube cannot represent all of tree space, it is a great improvement over a single summary tree, and it opens up exciting new opportunities for scaling up Bayesian phylogenetic inference. An open source implementation of the cubeVB algorithm is available from https://github.com/rbouckaert/cubevb as the cubevb package to BEAST 2.

show abstract

Section: Discussionmentioning

confidence: 85%

“…The more samples, the more accurate the matrix estimate but the longer it takes for the algorithm to finish. In that sense, it nicely fits the class of online algorithms (Bouckaert et al, 2022) that at any time can provide an answer, but the answer becomes more accurate the longer the algorithm runs. Detecting when longer runs do not increase accuracy of the posterior any more is an open question where automated convergence techniques employed for standard MCMC (Berling et al, 2023) could be exploited.…”

Section: /17mentioning

confidence: 79%

Variational Bayesian Phylogenies through Matrix Representation of Tree Space

Bouckaert

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…This framework is based on real-time genomic surveillance coupled with Bayesian phylogenetic inference. Recent computational advancements - such as the BICEPS, ORC, and online packages for BEAST 2 [22,23,24] - have made rapid Bayesian phylogenetic inference on large genomic datasets more feasible.…”

Section: Discussionmentioning

confidence: 99%

Tracing the international arrivals of SARS-CoV-2 Omicron variants after Aotearoa New Zealand reopened its border

Douglas

Winter

Ren

et al. 2022

Preprint

View full text Add to dashboard Cite

Recently there has been a surge in emergent SARS-CoV-2 lineages that are able to evade both vaccine induced immunity as well as prior infection from the founding Omicron BA.1 and BA.2 lineages. These highly transmissible and evasive lineages are on the rise and include Omicron variants BA.2.12.1, BA.4, and BA.5. Aotearoa New Zealand recently reopened its borders to many travellers, without their need to enter quarantine. By generating 10,403 complete SARS-CoV-2 genomes classified as Omicron, we show that New Zealand is observing an influx of these immune-evasive variants through the border. Specifically, there has been a recent surge of BA.5 and BA.2.12.1 introductions into the community and these can be explained by the gradual return to pre-pandemic levels of international traveller arrival rates. We estimate there is one Omicron transmission event from the border to the community for every ~5,000 passenger arrivals into the country, or around one introduction event per day at the current levels of travel. Given the waning levels of population immunity, this rate of importation presents the risk of a large wave in New Zealand during the second half of 2022. Genomic surveillance, coupled with modelling the rate at which new variants cross the border into the community, provides a lens on the rate at which new variants might gain a foothold and trigger new waves of infection.

show abstract

“…BEAST2 version 2.5.1 [69, 70] was used to construct a tip-dated phylogeny using core SNPs detected among all 219 genomes as input (see section “Variant calling and temporal diagnostics” above), an initial clock rate of 2.79×10 -7 substitutions/site/year [13], and an ascertainment bias correction to account for the use of solely variant sites [71]. bmodeltest [72] was used to infer a substitution model using Bayesian model averaging, with transitions and transversions split.…”

Section: Methodsmentioning

confidence: 99%

A multidrug-resistantSalmonella entericaTyphimurium DT104 complex lineage circulating among humans and cattle in the United States lost the ability to produce pertussis-like toxin ArtAB

Carroll

Piacenza²,

Cheng

et al. 2022

Preprint

View full text Add to dashboard Cite

Salmonella enterica subspecies enterica serotype Typhimurium phage type DT104 (DT104) can infect both humans and animals and is often multidrug-resistant (MDR). Previous studies have indicated that, unlike most S. Typhimurium, the overwhelming majority of DT104 strains produce a pertussis-like toxin, ArtAB, via prophage-encoded artAB; however, DT104 that lack artAB have been described on occasion. Here, we identify a MDR DT104 lineage circulating among humans and cattle in the United States, which lacks artAB (i.e., the “U.S. artAB-negative major clade”; n = 41 genomes). Unlike most other bovine- and human-associated DT104 strains from the U.S. (n = 219 total genomes), which harbor artAB on prophage Gifsy-1 (n = 167), members of the U.S. artAB-negative major clade lack Gifsy-1, as well as anti-inflammatory effector gogB. The U.S. artAB-negative major clade was predicted to have lost artAB, Gifsy-1, and gogB circa 1993-1995 (95% highest posterior density interval 1989.5-1997.7), in close temporal proximity to a predicted rapid increase in the U.S. DT104 effective population size (circa 1996). When compared to DT104 genomes from other world regions (n = 752 total genomes), several additional, sporadic artAB, Gifsy-1, and/or gogB loss events among clades encompassing ≤5 genomes were observed. In phenotypic assays that simulate conditions encountered during human and/or bovine digestion, members of the U.S. artAB-negative major clade did not differ from closely related Gifsy-1/artAB/gogB-harboring U.S. DT104 strains (ANOVA raw P > 0.05); thus, future research is needed to elucidate the roles that artAB, gogB, and Gifsy-1 play in DT104 virulence in humans and animals.Impact StatementMulti-drug resistant (MDR) Salmonella enterica serotype Typhimurium phage type DT104 (DT104) was responsible for a global epidemic among humans and animals throughout the 1990s and continues to circulate worldwide. Previous studies have indicated that the vast majority of DT104 produce a pertussis-like toxin, ArtAB, via prophage-encoded artAB. However, here we identify a DT104 lineage circulating among cattle and humans across ≥10 U.S. states, which lacks the ability to produce ArtAB (i.e., the “U.S. artAB-negative major clade”). The common ancestor of all U.S. artAB-negative major clade members lost the ability to produce ArtAB in close temporal proximity to the global MDR DT104 epidemic; however, the reason for this loss-of-function event within this well-established pathogen remains unclear. The role that ArtAB plays in DT104 virulence remains elusive, and phenotypic assays conducted here indicate that members of the U.S. artAB-negative major clade do not have a significant advantage or disadvantage relative to closely related Gifsy-1/artAB/gogB-harboring U.S. DT104 strains when exposed to stressors encountered during human and/or bovine digestion in vitro. However, ArtAB heterogeneity among DT104 suggest clade-specific selection for or against maintenance of ArtAB. Thus, future studies querying the virulence potential of the U.S. artAB- negative major clade are needed.Data SummaryThe authors confirm all supporting data, code and protocols have been provided within the article or through supplementary data files.

show abstract

Online Bayesian Analysis with BEAST 2

Cited by 13 publications

References 46 publications

Variational Bayesian Phylogenies through Matrix Representation of Tree Space

Variational Bayesian Phylogenies through Matrix Representation of Tree Space

Tracing the international arrivals of SARS-CoV-2 Omicron variants after Aotearoa New Zealand reopened its border

A multidrug-resistantSalmonella entericaTyphimurium DT104 complex lineage circulating among humans and cattle in the United States lost the ability to produce pertussis-like toxin ArtAB

Contact Info

Product

Resources

About