Bengt Sennblad

Simultaneous Bayesian gene tree reconstruction and reconciliation analysis

We present GSR, a probabilistic model integrating gene duplication, sequence evolution, and a relaxed molecular clock for substitution rates, that enables genomewide analysis of gene families. The gene duplication and loss process is a major cause for incongruence between gene and species tree, and deterministic methods have been developed to explain such differences through tree reconciliations. Although probabilistic methods for phylogenetic inference have been around for decades, probabilistic reconciliation methods are far less established. Based on our model, we have implemented a Bayesian analysis tool, PrIME-GSR, for gene tree inference that takes a known species tree into account. Our implementation is sound and we demonstrate its utility for genomewide gene-family analysis by applying it to recently presented yeast data. We validate PrIME-GSR by comparing with previous analyses of these data that take advantage of gene order information. In a case study we apply our method to the ADH gene family and are able to draw biologically relevant conclusions concerning gene duplications creating key yeast phenotypes. On a higher level this shows the biological relevance of our method. The obtained results demonstrate the value of a relaxed molecular clock. Our good performance will extend to species where gene order conservation is insufficient.

Gene tree reconstruction and orthology analysis based on an integrated model for duplications and sequence evolution

Gene tree and species tree reconstruction, orthology analysis and reconciliation, are problems important in multigenome-based comparative genomics and biology in general. In the present paper, we advance the frontier of these areas in several respects and provide important computational tools. First, exact algorithms are given for several probabilistic reconciliation problems with respect to the probabilistic gene evolution model, previously developed by the authors. Until now, those problems were solved by MCMC estimation algorithms. Second, we extend the gene evolution model to the gene sequence evolution model, by including sequence evolution. Third, we develop MCMC algorithms for the gene sequence evolution model that, given gene sequence data allows: (1) orthology analysis, reconciliation analysis, and gene tree reconstruction, w.r.t. a species tree, that balances a likely/unlikely reconciliation and a likely/unlikely gene tree and (2) species tree reconstruction that balance a likely/unlikely reconciliation and a likely/unlikely gene trees. These MCMC algorithms take advantage of the exact algorithms for the gene evolution model. We have successfully tested our dynamical programming algorithms on real data for a biogeography problem. The MCMC algorithms perform very well both on synthetic and biological data.

The gene evolution model and computing its associated probabilities

Phylogeny is both a fundamental tool in biology and a rich source of fascinating modeling and algorithmic problems. Todays wealth of sequenced genomes makes it increasingly important to understand evolutionary events such as duplications, losses, transpositions, inversions, lateral transfers, and domain shuffling. We focus on the gene duplication event, that constitutes a major force in the creation of genes with new function [Ohno 1970; Lynch and Force 2000] and, thereby also, of biodiversity. We introduce the probabilistic gene evolution model, which describes how a gene tree evolves within a given species tree with respect to speciation, gene duplication, and gene loss. The actual relation between gene tree and species tree is captured by a reconciliation, a concept which we generalize for more expressiveness. The model is a canonical generalization of the classical linear birth-death process, obtained by replacing the interval where the process takes place by a tree. For the gene evolution model, we derive efficient algorithms for some associated probability distributions: the probability of a reconciled tree, the probability of a gene tree, the maximum probability reconciliation, the posterior probability of a reconciliation, and sampling reconciliations with respect to the posterior probability. These algorithms provides the basis for several applications, including species tree construction, reconciliation analysis, orthology analysis, biogeography, and host-parasite co-evolution.

Sex-Specific Effects of Adiponectin on Carotid Intima-Media Thickness and Incident Cardiovascular Disease

Background Plasma adiponectin levels have previously been inversely associated with carotid intima-media thickness (IMT), a marker of subclinical atherosclerosis. In this study, we used a sex-stratified Mendelian randomization approach to investigate whether adiponectin has a causal protective influence on IMT. Methods and Results Baseline plasma adiponectin concentration was tested for association with baseline IMT, IMT progression over 30 months, and occurrence of cardiovascular events within 3 years in 3430 participants (women, n =1777; men, n =1653) with high cardiovascular risk but no prevalent disease. Plasma adiponectin levels were inversely associated with baseline mean bifurcation IMT after adjustment for established risk factors (β =−0.018, P<0.001) in men but not in women (β =−0.006, P =0.185; P for interaction =0.061). Adiponectin levels were inversely associated with progression of mean common carotid IMT in men (β =−0.0022, P =0.047), whereas no association was seen in women (0.0007, P =0.475; P for interaction =0.018). Moreover, we observed that adiponectin levels were inversely associated with coronary events in women (hazard ratio 0.57, 95% CI 0.37 to 0.87) but not in men (hazard ratio 0.82, 95% CI 0.54 to 1.25). A gene score of adiponectin-raising alleles in 6 loci, reported recently in a large multi-ethnic meta-analysis, was inversely associated with baseline mean bifurcation IMT in men (β =−0.0008, P =0.004) but not in women (β =−0.0003, P =0.522; P for interaction =0.007). Conclusions This report provides some evidence for adiponectin protecting against atherosclerosis, with effects being confined to men; however, compared with established cardiovascular risk factors, the effect of plasma adiponectin was modest. Further investigation involving mechanistic studies is warranted.

A phylogenetic analysis of 100+ genera and 50+ families of euasterids based on morphological and molecular data with notes on possible higher level morphological synapomorphies

Abstract. A data matrix of 143 morphological and chemical characters for 142 genera of euasterids according to the APG system was compiled and complemented with rbcL and ndhF sequences for most of the genera. The data were subjected to parsimony analysis and support was assessed by bootstrapping. Strict consensus trees from analyses of morphology alone and morphology + rbcL + ndhF are presented. The morphological data recover several groups supported by molecular data but at the level of orders and above relationships are only superficially in agreement with molecular studies. The analyses provide support for monophyly of Gentianales, Aquifoliales, Apiales, Asterales, and Dipsacales. All data indicate that Adoxaceae are closely related to Dipsacales and hence they should be included in that order. The trees were used to assess some possible morphological synapomorphies for euasterids I and II and for the orders of the APG system. Euasterids I are generally characterised by opposite leaves, entire leaf margins, hypogynous flowers, “early sympetaly” with a ring-shaped corolla primordium, fusion of stamen filaments with the corolla tube, and capsular fruits. Euasterids II often have alternate leaves, serrate-dentate leaf margins, epigynous flowers, “late sympetaly” with distinct petal primordia, free stamen filaments, and indehiscent fruits. It is unclear which of these characters represent synapomorphies and symplesiomorphies for the two groups, respectively, and there are numerous expections to be interpreted as reversals and parallelisms.

Genome-wide association study for circulating levels of PAI-1 provides novel insights into its regulation.

We conducted a genome-wide association study to identify novel associations between genetic variants and circulating plasminogen activator inhibitor-1 (PAI-1) concentration, and examined functional implications of variants and genes that were discovered. A discovery meta-analysis was performed in 19 599 subjects, followed by replication analysis of genome-wide significant (P < 5 × 10(-8)) single nucleotide polymorphisms (SNPs) in 10 796 independent samples. We further examined associations with type 2 diabetes and coronary artery disease, assessed the functional significance of the SNPs for gene expression in human tissues, and conducted RNA-silencing experiments for one novel association. We confirmed the association of the 4G/5G proxy SNP rs2227631 in the promoter region of SERPINE1 (7q22.1) and discovered genome-wide significant associations at 3 additional loci: chromosome 7q22.1 close to SERPINE1 (rs6976053, discovery P = 3.4 × 10(-10)); chromosome 11p15.2 within ARNTL (rs6486122, discovery P = 3.0 × 10(-8)); and chromosome 3p25.2 within PPARG (rs11128603, discovery P = 2.9 × 10(-8)). Replication was achieved for the 7q22.1 and 11p15.2 loci. There was nominal association with type 2 diabetes and coronary artery disease at ARNTL (P < .05). Functional studies identified MUC3 as a candidate gene for the second association signal on 7q22.1. In summary, SNPs in SERPINE1 and ARNTL and an SNP associated with the expression of MUC3 were robustly associated with circulating levels of PAI-1.

Probabilistic Orthology Analysis

Orthology analysis aims at identifying orthologous genes and gene products from different organisms and, therefore, is a powerful tool in modern computational and experimental biology. Although reconciliation-based orthology methods are generally considered more accurate than distance-based ones, the traditional parsimony-based implementation of reconciliation-based orthology analysis (most parsimonious reconciliation [MPR]) suffers from a number of shortcomings. For example, 1) it is limited to orthology predictions from the reconciliation that minimizes the number of gene duplication and loss events, 2) it cannot evaluate the support of this reconciliation in relation to the other reconciliations, and 3) it cannot make use of prior knowledge (e.g., about species divergence times) that provides auxiliary information for orthology predictions. We present a probabilistic approach to reconciliation-based orthology analysis that addresses all these issues by estimating orthology probabilities. The method is based on the gene evolution model, an explicit evolutionary model for gene duplication and gene loss inside a species tree, that generalizes the standard birth-death process. We describe the probabilistic approach to orthology analysis using 2 experimental data sets and show that the use of orthology probabilities allows a more informative analysis than MPR and, in particular, that it is less sensitive to taxon sampling problems. We generalize these anecdotal observations and show, using data generated under biologically realistic conditions, that MPR give false orthology predictions at a substantial frequency. Last, we provide a new orthology prediction method that allows an orthology and paralogy classification with any chosen sensitivity/specificity combination from the spectra of achievable combinations. We conclude that probabilistic orthology analysis is a strong and more advanced alternative to traditional orthology analysis and that it provides a framework for sophisticated comparative studies of processes in genome evolution.

A Bayesian Method for Analyzing Lateral Gene Transfer

Lateral gene transfer (LGT)--which transfers DNA between two non-vertically related individuals belonging to the same or different species--is recognized as a major force in prokaryotic evolution, and evidence of its impact on eukaryotic evolution is ever increasing. LGT has attracted much public attention for its potential to transfer pathogenic elements and antibiotic resistance in bacteria, and to transfer pesticide resistance from genetically modified crops to other plants. In a wider perspective, there is a growing body of studies highlighting the role of LGT in enabling organisms to occupy new niches or adapt to environmental changes. The challenge LGT poses to the standard tree-based conception of evolution is also being debated. Studies of LGT have, however, been severely limited by a lack of computational tools. The best currently available LGT algorithms are parsimony-based phylogenetic methods, which require a pre-computed gene tree and cannot choose between sometimes wildly differing most parsimonious solutions. Moreover, in many studies, simple heuristics are applied that can only handle putative orthologs and completely disregard gene duplications (GDs). Consequently, proposed LGT among specific gene families, and the rate of LGT in general, remain debated. We present a Bayesian Markov-chain Monte Carlo-based method that integrates GD, gene loss, LGT, and sequence evolution, and apply the method in a genome-wide analysis of two groups of bacteria: Mollicutes and Cyanobacteria. Our analyses show that although the LGT rate between distant species is high, the net combined rate of duplication and close-species LGT is on average higher. We also show that the common practice of disregarding reconcilability in gene tree inference overestimates the number of LGT and duplication events.

Serum 25-Hydroxyvitamin D Concentration in Subclinical Carotid Atherosclerosis

Objective—Vitamin D deficiency has been implicated in cardiovascular disease and is associated with multiple cardiovascular risk factors. We investigated the serum 25-hydroxyvitamin D (25(OH)D) concentration in relation to latitude, baseline carotid intima-media thickness (IMT), and IMT progression, the carotid IMT measures being surrogate markers of subclinical atherosclerosis and cardiovascular disease risk. Approach and Results—Serum 25(OH)D concentration was related to high-resolution carotid IMT measures in 3430 middle-aged and elderly subjects with high cardiovascular risk but no prevalent disease, who were recruited at 7 centers in Finland, Sweden, The Netherlands, France, and Italy. Participants underwent carotid ultrasound examination at baseline and at months 15 and 30 after entry into the study, whereas blood samples, clinical data, and information about lifestyle were collected at baseline. Serum 25(OH)D levels were positively associated with latitude (Jonckheere–Terpstra &khgr;=166.643; P<0.001) and, as previously reported, associated with a range of cardiovascular risk factors. There were no independent relationships between 25(OH)D and segment-specific or composite IMT measures in the entire cohort. In analyses stratified by sex, diabetes mellitus, and statin treatment, weak associations with some baseline and progression measures of carotid IMT were observed in males, diabetics, and nonstatin-treated individuals. Conclusions—Levels of 25(OH)D differed across Europe, were highest in the North, showed multiple associations with established and emerging cardiovascular risk factors but were not consistently, independently related to measures of carotid IMT. This argues against a protective role of vitamin D against subclinical atherosclerosis in high-risk individuals.

Evaluation of Bayesian Models of Substitution Rate Evolution—Parental Guidance versus Mutual Independence

We have evaluated the performance of two classes of probabilistic models for substitution rate variation over phylogenetic trees. In the first class, branch rates are considered to be independent and identically distributed (i.i.d.) stochastic variables. Three versions with respect to the underlying distribution (Gamma, Inverse Gaussian, and LogNormal) are considered. The i.i.d. models are compared with the autocorrelated (AC) model, where rates of adjacent nodes in the tree are AC, so that a node rate is LogNormal distributed around the rate of the parent node. The performance of different models is evaluated using three empirical data sets. For all data sets, it was clear that all tested models extracted substantial knowledge from data when posterior divergence time distributions were compared with the prior distributions and, furthermore, that they clearly outperformed a molecular clock. Moreover, the descriptive power of the i.i.d. models, as evaluated by Bayes factors, was either equal to or clearly better than that of the AC model. The latter effect increased with extended taxon sampling. Likewise, under none of the models could we find compelling evidence, in any of the data sets, for rate correlation between adjacent branches/nodes. These findings challenge previous suggestions of universality of autocorrelation in sequence evolution. We also performed an additional comparison with a divergence time prior including calibration information from fossil evidence. Adding fossil information to the prior had negligible effect on Bayes factors and mainly affected the width of the posterior distribution of the divergence times, whereas the relative position of the mean divergence times were largely unaffected.