A Nonparametric Bayesian Model for Local Clustering with Application to Proteomics

Clustering computer mouse tracking data with informed hierarchical shrinkage partition priors

Mouse-tracking data, which record computer mouse trajectories while participants perform an experimental task, provide valuable insights into subjects' underlying cognitive processes. Neuroscientists are interested in clustering the subjects' responses during computer mouse-tracking tasks to reveal patterns of individual decision-making behaviors and identify population subgroups with similar neurobehavioral responses. These data can be combined with neuroimaging data to provide additional information for personalized interventions. In this article, we develop a novel hierarchical shrinkage partition (HSP) prior for clustering summary statistics derived from the trajectories of mouse-tracking data. The HSP model defines a subjects' cluster as a set of subjects that gives rise to more similar (rather than identical) nested partitions of the conditions. The proposed model can incorporate prior information about the partitioning of either subjects or conditions to facilitate clustering, and it allows for deviations of the nested partitions within each subject group. These features distinguish the HSP model from other bi-clustering methods that typically create identical nested partitions of conditions within a subject group. Furthermore, it differs from existing nested clustering methods, which define clusters based on common parameters in the sampling model and identify subject groups by different distributions. We illustrate the unique features of the HSP model on a mouse tracking dataset from a pilot study and in simulation studies. Our results show the ability and effectiveness of the proposed exploratory framework in clustering and revealing possible different behavioral patterns across subject groups.

Proteome trait regulation of marine Synechococcus elemental stoichiometry under global change

Recent studies have demonstrated regional differences in marine ecosystem C:N:P with implications for carbon and nutrient cycles. Due to strong co-variance, temperature and nutrient stress explain variability in C:N:P equally well. A reductionistic approach can link changes in individual environmental drivers with changes in biochemical traits and cell C:N:P. Thus, we quantified effects of temperature and nutrient stress on Synechococcus chemistry using laboratory chemostats, chemical analyses, and data-independent acquisition mass spectrometry proteomics. Nutrient supply accounted for most C:N:Pcell variability and induced tradeoffs between nutrient acquisition and ribosomal proteins. High temperature prompted heat-shock, whereas thermal effects via the "translation-compensation hypothesis" were only seen under P-stress. A Nonparametric Bayesian Local Clustering algorithm suggested that changes in lipopolysaccharides, peptidoglycans, and C-rich compatible solutes may also contribute to C:N:P regulation. Physiological responses match field-based trends in ecosystem stoichiometry and suggest a hierarchical environmental regulation of current and future ocean C:N:P.

BAREB: A Bayesian repulsive biclustering model for periodontal data

Preventing periodontal diseases (PD) and maintaining the structure and function of teeth are important goals for personal oral care. To understand the heterogeneity in patients with diverse PD patterns, we develop a Bayesian repulsive biclustering method that can simultaneously cluster the PD patients and their tooth sites after taking the patient- and site-level covariates into consideration. BAREB uses the determinantal point process prior to induce diversity among different biclusters to facilitate parsimony and interpretability. Since PD progression is hypothesized to be spatially referenced, BAREB factors in the spatial dependence among tooth sites. In addition, since PD is the leading cause for tooth loss, the missing data mechanism is nonignorable. Such nonrandom missingness is incorporated into BAREB. For the posterior inference, we design an efficient reversible jump Markov chain Monte Carlo sampler. Simulation studies show that BAREB is able to accurately estimate the biclusters, and compares favorably to alternatives. For real world application, we apply BAREB to a dataset from a clinical PD study, and obtain desirable and interpretable results. A major contribution of this article is the Rcpp implementation of our methodology, available in the R package BAREB.

Sample selection in the face of design constraints: Use of clustering to define sample strata for qualitative research

OBJECTIVE

To sample 40 physician organizations stratified on the basis of longitudinal cost of care measures for qualitative interviews in order to describe the range of care delivery structures and processes that are being deployed to influence the total costs of caring for patients.

DATA SOURCES

Three years of physician organization-level total cost of care data (n = 156 in California) from the Integrated Healthcare Association's value-based pay-for-performance program.

STUDY DESIGN

We fit total cost of care data using mixture and K-means clustering algorithms to segment the population of physician organizations into sampling strata based on 3-year cost trajectories (ie, cost curves).

PRINCIPAL FINDINGS

A mixture of multivariate normal distributions can classify physician organization cost curves into clusters defined by total cost level, shape, and within-cluster variation. K-means clustering does not accommodate differing levels of within-cluster variation and resulted in more clusters being allocated to unstable cost curves. A mixture of regressions approach focuses overly on anomalous trajectories and is sensitive to model coding.

CONCLUSIONS

Statistical clustering can be used to form sampling strata when longitudinal measures are of primary interest. Many clustering algorithms are available; the choice of the clustering algorithm can strongly impact the resulting strata because various algorithms focus on different aspects of the observed data.

Joint clustering with correlated variables

Traditional clustering methods focus on grouping subjects or (dependent) variables assuming independence between the variables. Clusters formed through these approaches can potentially lack homogeneity. This article proposes a joint clustering method by which both variables and subjects are clustered. In each joint cluster (in general composed of a subset of variables and a subset of subjects), there exists a unique association between dependent variables and covariates of interest. To this end, a Bayesian method is designed, in which a semi-parametric model is used to evaluate any unknown relationships between possibly correlated variables and covariates of interest, and a Dirichlet process is utilized to cluster subjects. Compared to existing clustering techniques, the major novelty of the method exists in its ability to improve the homogeneity of clusters, along with the ability to take the correlations between variables into account. Via simulations, we examine the performance and efficiency of the proposed method. Applying the method to cluster allergens and subjects based on the association of wheal size in reaction to allergens with age, we found that a certain pattern of allergic sensitization to a set of allergens has a potential to reduce the occurrence of asthma.

Clustering distributions with the marginalized nested Dirichlet process

We introduce a marginal version of the nested Dirichlet process to cluster distributions or histograms. We apply the model to cluster genes by patterns of gene-gene interaction. The proposed approach is based on the nested partition that is implied in the original construction of the nested Dirichlet process. It allows simulation exact inference, as opposed to a truncated Dirichlet process approximation. More importantly, the construction highlights the nature of the nested Dirichlet process as a nested partition of experimental units. We apply the proposed model to inference on clustering genes related to DNA mismatch repair (DMR) by the distribution of gene-gene interactions with other genes. Gene-gene interactions are recorded as coefficients in an auto-logistic model for the co-expression of two genes, adjusting for copy number variation, methylation and protein activation. These coefficients are extracted from an online database, called Zodiac, computed based on The Cancer Genome Atlas (TCGA) data. We compare results with a variation of k-means clustering that is set up to cluster distributions, truncated NDP and a hierarchical clustering method. The proposed inference shows favorable performance, under simulated conditions and also in the real data sets.

Strategies for Advancing Disease Definition Using Biomarkers and Genetics: The Bipolar and Schizophrenia Network for Intermediate Phenotypes

It is critical for psychiatry as a field to develop approaches to define the molecular, cellular, and circuit basis of its brain diseases, especially for serious mental illnesses, and then to use these definitions to generate biologically based disease categories, as well as to explore disease mechanisms and illness etiologies. Our current reliance on phenomenology is inadequate to support exploration of molecular treatment targets and disease formulations, and the leap directly from phenomenology to disease biology has been limiting because of broad heterogeneity within conventional diagnoses. The questions addressed in this review are formulated around how we can use brain biomarkers to achieve disease categories that are biologically based. We have grouped together a series of vignettes as examples of early approaches, all using the Bipolar and Schizophrenia Network on Intermediate Phenotypes (BSNIP) biomarker database and collaborators, starting off with describing the foundational statistical methods for these goals. We use primarily criterion-free statistics to identify pertinent groups of involved genes related to psychosis as well as symptoms, and finally, to create new biologically based disease cohorts within the psychopathological dimension of psychosis. Although we do not put these results forward as final formulations, they represent a novel effort to rely minimally on phenomenology as a diagnostic tool and to fully embrace brain characteristics of structure, as well as molecular and cellular characteristics and function, to support disease definition in psychosis.

A Nonparametric Bayesian Model for Nested Clustering

We propose a nonparametric Bayesian model for clustering where clusters of experimental units are determined by a shared pattern of clustering another set of experimental units. The proposed model is motivated by the analysis of protein activation data, where we cluster proteins such that all proteins in one cluster give rise to the same clustering of patients. That is, we define clusters of proteins by the way that patients group with respect to the corresponding protein activations. This is in contrast to (almost) all currently available models that use shared parameters in the sampling model to define clusters. This includes in particular model based clustering, Dirichlet process mixtures, product partition models, and more. We show results for two typical biostatistical inference problems that give rise to clustering.

Nonparametric Bayesian Bi-Clustering for Next Generation Sequencing Count Data

Histone modifications (HMs) play important roles in transcription through post-translational modifications. Combinations of HMs, known as chromatin signatures, encode specific messages for gene regulation. We therefore expect that inference on possible clustering of HMs and an annotation of genomic locations on the basis of such clustering can contribute new insights about the functions of regulatory elements and their relationships to combinations of HMs. We propose a nonparametric Bayesian local clustering Poisson model (NoB-LCP) to facilitate posterior inference on two-dimensional clustering of HMs and genomic locations. The NoB-LCP clusters HMs into HM sets and lets each HM set define its own clustering of genomic locations. Furthermore, it probabilistically excludes HMs and genomic locations that are irrelevant to clustering. By doing so, the proposed model effectively identifies important sets of HMs and groups regulatory elements with similar functionality based on HM patterns.