Marko Järvenpää

Bayesian analysis of GUHA hypotheses

The LISp-Miner system for data mining and knowledge discovery uses the GUHA method to comb through a large data base and finds 2 × 2 contingency tables that satisfy a certain condition given by generalised quantifiers and thereby suggest the existence of possible relations between attributes. In this paper, we show how a more detailed interpretation of the data in the tables that were found by GUHA can be obtained using Bayesian statistical methods. Using a multinomial sampling model and Dirichlet prior, we derive posterior distributions for parameters that correspond to GUHA generalised quantifiers. Examples are presented illustrating the new Bayesian post-processing tools implemented in LISp-Miner. A statistical model for the analysis of contingency tables for data from two subpopulations is also presented.

A Bayesian model of acquisition and clearance of bacterial colonization incorporating within-host variation

Bacterial populations that colonize a host can play important roles in host health, including serving as a reservoir that transmits to other hosts and from which invasive strains emerge, thus emphasizing the importance of understanding rates of acquisition and clearance of colonizing populations. Studies of colonization dynamics have been based on assessment of whether serial samples represent a single population or distinct colonization events. With the use of whole genome sequencing to determine genetic distance between isolates, a common solution to estimate acquisition and clearance rates has been to assume a fixed genetic distance threshold below which isolates are considered to represent the same strain. However, this approach is often inadequate to account for the diversity of the underlying within-host evolving population, the time intervals between consecutive measurements, and the uncertainty in the estimated acquisition and clearance rates. Here, we present a fully Bayesian model that provides probabilities of whether two strains should be considered the same, allowing us to determine bacterial clearance and acquisition from genomes sampled over time. Our method explicitly models the within-host variation using population genetic simulation, and the inference is done using a combination of Approximate Bayesian Computation (ABC) and Markov Chain Monte Carlo (MCMC). We validate the method with multiple carefully conducted simulations and demonstrate its use in practice by analyzing a collection of methicillin resistant Staphylococcus aureus (MRSA) isolates from a large recently completed longitudinal clinical study. An R-code implementation of the method is freely available at: https://github.com/mjarvenpaa/bacterial-colonization-model.git. Author summary As colonizing bacterial populations are the source for much transmission and a reservoir for infection, they are a major focus of interest clinically and epidemiologically. Understanding the dynamics of colonization depends on being able to confidently identify acquisition and clearance events given intermittent sampling of hosts. To do so, we need a model of within-host bacterial population evolution from acquisition through the time of sampling that enables estimation of whether two samples are derived from the same population. Past efforts have frequently relied on empirical genetic distance thresholds that forgo an underlying model or employ a simple molecular clock model. Here, we present an inferential method that accounts for the timing of sample collection and population diversification, to provide a probabilistic estimate for whether two isolates represent the same colonizing strain. This method has implications for understanding the dynamics of acquisition and clearance of colonizing bacteria, and the impact on these rates by factors such as sensitivity of the sampling method, pathogen genotype, competition with other carriage bacteria, host immune response, and antibiotic exposure.

Efficient acquisition rules for model-based approximate Bayesian computation

Approximate Bayesian computation (ABC) is a method for Bayesian inference when the likelihood is unavailable but simulating from the model is possible. However, many ABC algorithms require a large number of simulations, which can be costly. To reduce the computational cost, Bayesian optimisation (BO) and surrogate models such as Gaussian processes have been proposed. Bayesian optimisation enables one to intelligently decide where to evaluate the model next but common BO strategies are not designed for the goal of estimating the posterior distribution. Our paper addresses this gap in the literature. We propose to compute the uncertainty in the ABC posterior density, which is due to a lack of simulations to estimate this quantity accurately, and define a loss function that measures this uncertainty. We then propose to select the next evaluation location to minimise the expected loss. Experiments show that the proposed method often produces the most accurate approximations as compared to common BO strategies.

A generator of morphological clones for plant species

Detailed and realistic tree form generators have numerous applications in ecology and forestry. Here, we present an algorithm for generating morphological tree “clones” based on the detailed reconstruction of the laser scanning data, statistical measure of similarity, and a plant growth algorithm with simple stochastic rules. The algorithm is designed to produce tree forms, i.e. morphological clones, similar as a whole (coarse-grain scale), but varying in minute details of organization (fine-grain scale). We present a general procedure for obtaining these morphological clones. Although we opted for certain choices in our algorithm, its various parts may vary depending on the application. Namely, we have shown that specific multi-purpose procedural stochastic growth model can be algorithmically adjusted to produce the morphological clones replicated from the target experimentally measured tree. For this, we have developed a statistical measure of similarity (structural distance) between any given pair of trees, which allows for the comprehensive comparing of the tree morphologies in question by means of empirical distributions describing geometrical and topological features of a tree. Our algorithm can be used in variety of applications and contexts for exploration of the morphological potential of the growth models, arising in all sectors of plant science research. Summary Statement We present an algorithmic framework, based on the Bayesian inference, for generating morphological tree clones using a combination of stochastic growth models and experimentally derived tree structures.

Bayes Forest: a data-intensive generator of morphological tree clones

Abstract Detailed and realistic tree form generators have numerous applications in ecology and forestry. For example, the varying morphology of trees contributes differently to formation of landscapes, natural habitats of species, and eco-physiological characteristics of the biosphere. Here, we present an algorithm for generating morphological tree “clones” based on the detailed reconstruction of the laser scanning data, statistical measure of similarity, and a plant growth model with simple stochastic rules. The algorithm is designed to produce tree forms, i.e., morphological clones, similar (and not identical) in respect to tree-level structure, but varying in fine-scale structural detail. Although we opted for certain choices in our algorithm, individual parts may vary depending on the application, making it a general adaptable pipeline. Namely, we showed that a specific multipurpose procedural stochastic growth model can be algorithmically adjusted to produce the morphological clones replicated from the target experimentally measured tree. For this, we developed a statistical measure of similarity (structural distance) between any given pair of trees, which allows for the comprehensive comparing of the tree morphologies by means of empirical distributions describing the geometrical and topological features of a tree. Finally, we developed a programmable interface to manipulate data required by the algorithm. Our algorithm can be used in a variety of applications for exploration of the morphological potential of the growth models (both theoretical and experimental), arising in all sectors of plant science research.