Detecting associations between intact connectomes and clinical covariates using recursive partitioning object-oriented data analysis.

Many neuroscientists are interested in how connectomes (graphical representations of functional connectivity between areas of the brain) change in relation to covariates. In statistics, changes like this are analyzed using regression, where the outcomes or dependent variables are regressed onto the covariates. However, when the outcome is a complex object, such as connectome graphs, classical regression models cannot be used. The regression approach developed here to work with complex graph outcomes combines recursive partitioning with the Gibbs distribution. We will only discuss the application to connectomes, but the method is generally applicable to any graphical outcome. The method, called Gibbs-RPart, partitions the covariate space into a set of nonoverlapping regions such that the connectomes within regions are more similar than they are to the connectomes in other regions. This paper extends the object-oriented data analysis paradigm for graph-valued data based on the Gibbs distribution, which we have applied previously to hypothesis testing to compare populations of connectomes from distinct groups (see the work of La Rosa et al).

Gut bacteria dysbiosis and necrotising enterocolitis in very low birthweight infants: a prospective case-control study.

BACKGROUND: Gut bacteria might predispose to or protect from necrotising enterocolitis, a severe illness linked to prematurity. In this observational prospective study we aimed to assess whether one or more bacterial taxa in the gut differ between infants who subsequently develop necrotising enterocolitis (cases) and those who do not (controls). METHODS: We enrolled very low birthweight (1500 g and lower) infants in the primary cohort (St Louis Children's Hospital) between July 7, 2009, and Sept 16, 2013, and in the secondary cohorts (Kosair Children's Hospital and Children's Hospital at Oklahoma University) between Sept 12, 2011 and May 25, 2013. We prospectively collected and then froze stool samples for all infants. Cases were defined as infants whose clinical courses were consistent with necrotising enterocolitis and whose radiographs fulfilled criteria for Bell's stage 2 or 3 necrotising enterocolitis. Control infants (one to four per case; not fixed ratios) with similar gestational ages, birthweight, and birth dates were selected from the population after cases were identified. Using primers specific for bacterial 16S rRNA genes, we amplified and then pyrosequenced faecal DNA from stool samples. With use of Dirichlet multinomial analysis and mixed models to account for repeated measures, we identified host factors, including development of necrotising enterocolitis, associated with gut bacterial populations. FINDINGS: We studied 2492 stool samples from 122 infants in the primary cohort, of whom 28 developed necrotising enterocolitis; 94 infants were used as controls. The microbial community structure in case stools differed significantly from those in control stools. These differences emerged only after the first month of age. In mixed models, the time-by-necrotising-enterocolitis interaction was positively associated with Gammaproteobacteria (p=0·0010) and negatively associated with strictly anaerobic bacteria, especially Negativicutes (p=0·0019). We studied 1094 stool samples from 44 infants in the secondary cohorts. 18 infants developed necrotising enterocolitis (cases) and 26 were controls. After combining data from all cohorts (166 infants, 3586 stools, 46 cases of necrotising enterocolitis), there were increased proportions of Gammaproteobacteria (p=0·0011) and lower proportions of both Negativicutes (p=0·0013) and the combined Clostridia-Negativicutes class (p=0·0051) in infants who went on to develop necrotising enterocolitis compared with controls. These associations were strongest in both the primary cohort and the overall cohort for infants born at less than 27 weeks' gestation. INTERPRETATION: A relative abundance of Gammaproteobacteria (ie, Gram-negative facultative bacilli) and relative paucity of strict anaerobic bacteria (especially Negativicutes) precede necrotising enterocolitis in very low birthweight infants. These data offer candidate targets for interventions to prevent necrotising enterocolitis, at least among infants born at less than 27 weeks' gestation. FUNDING: National Institutes of Health (NIH), Foundation for the NIH, the Children's Discovery Institute.

Patterned progression of bacterial populations in the premature infant gut.

In the weeks after birth, the gut acquires a nascent microbiome, and starts its transition to bacterial population equilibrium. This early-in-life microbial population quite likely influences later-in-life host biology. However, we know little about the governance of community development: does the gut serve as a passive incubator where the first organisms randomly encountered gain entry and predominate, or is there an orderly progression of members joining the community of bacteria? We used fine interval enumeration of microbes in stools from multiple subjects to answer this question. We demonstrate via 16S rRNA gene pyrosequencing of 922 specimens from 58 subjects that the gut microbiota of premature infants residing in a tightly controlled microbial environment progresses through a choreographed succession of bacterial classes from Bacilli to Gammaproteobacteria to Clostridia, interrupted by abrupt population changes. As infants approach 33-36 wk postconceptional age (corresponding to the third to the twelfth weeks of life depending on gestational age at birth), the gut is well colonized by anaerobes. Antibiotics, vaginal vs. Caesarian birth, diet, and age of the infants when sampled influence the pace, but not the sequence, of progression. Our results suggest that in infants in a microbiologically constrained ecosphere of a neonatal intensive care unit, gut bacterial communities have an overall nonrandom assembly that is punctuated by microbial population abruptions. The possibility that the pace of this assembly depends more on host biology (chiefly gestational age at birth) than identifiable exogenous factors warrants further consideration.

Gibbs distribution for statistical analysis of graphical data with a sample application to fcMRI brain images.

This paper develops object-oriented data analysis (OODA) statistical methods that are novel and complementary to existing methods of analysis of human brain scan connectomes, defined as graphs representing brain anatomical or functional connectivity. OODA is an emerging field where classical statistical approaches (e.g., hypothesis testing, regression, estimation, and confidence intervals) are applied to data objects such as graphs or functions. By analyzing data objects directly we avoid loss of information that occurs when data objects are transformed into numerical summary statistics. By providing statistical tools that analyze sets of connectomes without loss of information, new insights into neurology and medicine may be achieved. In this paper we derive the formula for statistical model fitting, regression, and mixture models; test their performance in simulation experiments; and apply them to connectomes from fMRI brain scans collected during a serial reaction time task study. Software for fitting graphical object-oriented data analysis is provided.

New statistical method identifies cytokines that distinguish stool microbiomes.

Regressing an outcome or dependent variable onto a set of input or independent variables allows the analyst to measure associations between the two so that changes in the outcome can be described by and predicted by changes in the inputs. While there are many ways of doing this in classical statistics, where the dependent variable has certain properties (e.g., a scalar, survival time, count), little progress on regression where the dependent variable are microbiome taxa counts has been made that do not impose extremely strict conditions on the data. In this paper, we propose and apply a new regression model combining the Dirichlet-multinomial distribution with recursive partitioning providing a fully non-parametric regression model. This model, called DM-RPart, is applied to cytokine data and microbiome taxa count data and is applicable to any microbiome taxa count/metadata, is automatically fit, and intuitively interpretable. This is a model which can be applied to any microbiome or other compositional data and software (R package HMP) available through the R CRAN website.

Statistical object data analysis of taxonomic trees from human microbiome data.

Human microbiome research characterizes the microbial content of samples from human habitats to learn how interactions between bacteria and their host might impact human health. In this work a novel parametric statistical inference method based on object-oriented data analysis (OODA) for analyzing HMP data is proposed. OODA is an emerging area of statistical inference where the goal is to apply statistical methods to objects such as functions, images, and graphs or trees. The data objects that pertain to this work are taxonomic trees of bacteria built from analysis of 16S rRNA gene sequences (e.g. using RDP); there is one such object for each biological sample analyzed. Our goal is to model and formally compare a set of trees. The contribution of our work is threefold: first, a weighted tree structure to analyze RDP data is introduced; second, using a probability measure to model a set of taxonomic trees, we introduce an approximate MLE procedure for estimating model parameters and we derive LRT statistics for comparing the distributions of two metagenomic populations; and third the Jumpstart HMP data is analyzed using the proposed model providing novel insights and future directions of analysis.