Background

Composite scales have recently been proposed as outcome measures for clinical trials. For example, the Prodromal Alzheimer's Cognitive Composite (PACC) is the sum of z-score normed component measures assessing episodic memory, timed executive function, and global cognition. Alternative methods of calculating composite total scores using the weighted sum of the component measures that maximize signal-to-noise of the resulting composite score have been proposed. Optimal weights can be estimated from pilot data, but it is an open question how large a pilot trial is required to calculate reliably optimal weights.

Methods

In this manuscript, we describe the calculation of optimal weights, and use large-scale computer simulations to investigate the question of how large a pilot study sample is required to inform the calculation of optimal weights. The simulations are informed by the pattern of decline observed in cognitively normal subjects enrolled in the Alzheimer's Disease Cooperative Study (ADCS) Prevention Instrument cohort study, restricting to n=75 subjects age 75 and over with an ApoE E4 risk allele and therefore likely to have an underlying Alzheimer neurodegenerative process.

Results

In the context of secondary prevention trials in Alzheimer's disease, and using the components of the PACC, we found that pilot studies as small as 100 are sufficient to meaningfully inform weighting parameters. Regardless of the pilot study sample size used to inform weights, the optimally weighted PACC consistently outperformed the standard PACC in terms of statistical power to detect treatment effects in a clinical trial. Pilot studies of size 300 produced weights that achieved near-optimal statistical power, and reduced required sample size relative to the standard PACC by more than half.

Conclusions

These simulations suggest that modestly sized pilot studies, comparable to that of a phase 2 clinical trial, are sufficient to inform the construction of composite outcome measures. Although these findings apply only to the PACC in the context of prodromal AD, the observation that weights only have to approximate the optimal weights to achieve near-optimal performance should generalize. Performing a pilot study or phase 2 trial to inform the weighting of proposed composite outcome measures is highly cost-effective. The net effect of more efficient outcome measures is that smaller trials will be required to test novel treatments. Alternatively, second generation trials can use prior clinical trial data to inform weighting, so that greater efficiency can be achieved as we move forward.

Low concordance between studies that examine the role of microbiota in human diseases is a pervasive challenge that limits the capacity to identify causal relationships between host-associated microorganisms and pathology. The risk of obtaining false positives is exacerbated by wide interindividual heterogeneity in microbiota composition¹, probably due to population-wide differences in human lifestyle and physiological variables² that exert differential effects on the microbiota. Here we infer the greatest, generalized sources of heterogeneity in human gut microbiota profiles and also identify human lifestyle and physiological characteristics that, if not evenly matched between cases and controls, confound microbiota analyses to produce spurious microbial associations with human diseases. We identify alcohol consumption frequency and bowel movement quality as unexpectedly strong sources of gut microbiota variance that differ in distribution between healthy participants and participants with a disease and that can confound study designs. We demonstrate that for numerous prevalent, high-burden human diseases, matching cases and controls for confounding variables reduces observed differences in the microbiota and the incidence of spurious associations. On this basis, we present a list of host variables that we recommend should be captured in human microbiota studies for the purpose of matching comparison groups, which we anticipate will increase robustness and reproducibility in resolving the members of the gut microbiota that are truly associated with human disease.

BACKGROUND: Growing evidence suggests bidirectional links between gut microbiota and sleep quality as shared contributors to health. Little is known about the relationship between microbiota and sleep among older persons. METHODS: We used 16S rRNA sequencing to characterize stool microbiota among men (n = 606, mean [standard deviation] age = 83.9 [3.8]) enrolled in the Osteoporotic Fractures in Men (MrOS) study from 2014 to 2016. Sleep was assessed concurrently by a questionnaire (Pittsburgh Sleep Quality index [PSQI]), and activity monitor to examine timing (acrophase) and regularity of patterns (F-statistic). Alpha diversity was measured using Faiths phylogenetic diversity (PD). Beta diversity was calculated with robust Aitchison distance with matrix completion (RPCA) and phylogenetic-RPCA (PRPCA). Their association with sleep variables was tested with partial distance-based redundancy analysis (dbRDA). Predictive-ratio biomarkers associated with sleep measurements were identified with CoDaCoRe. RESULTS: In unadjusted analyses, men with poor sleep (PSQI >5) tended to have lower alpha diversity compared to men with normal sleep (Faiths PD, beta = -0.15; 95% confidence interval [CI]: -0.30 to 0.01, p = .06). Sleep regularity was significantly associated with RPCA and PRPCA, even after adjusting for site, batch, age, ethnicity, body mass index, diabetes, antidepressant and sleep medication use, and health behaviors (RPCA/PRPCA dbRDA; p = .033/.002). In taxonomic analysis, ratios of 7:6 bacteria for better regularity (p = .0004) and 4:7 for worse self-reported sleep (p = .005) were differentially abundant: some butyrate-producing bacteria were associated with better sleep characteristics. CONCLUSIONS: Subjective and objective indicators of sleep quality suggest that older men with better sleep patterns are more likely to harbor butyrate-producing bacteria associated with better health.

Differential abundance testing is a critical task in microbiome studies that is complicated by the sparsity of data matrices. Here we adapt for microbiome studies a solution from the field of gene expression analysis to produce a new method, discrete false-discovery rate (DS-FDR), that greatly improves the power to detect differential taxa by exploiting the discreteness of the data. Additionally, DS-FDR is relatively robust to the number of noninformative features, and thus removes the problem of filtering taxonomy tables by an arbitrary abundance threshold. We show by using a combination of simulations and reanalysis of nine real-world microbiome data sets that this new method outperforms existing methods at the differential abundance testing task, producing a false-discovery rate that is up to threefold more accurate, and halves the number of samples required to find a given difference (thus increasing the efficiency of microbiome experiments considerably). We therefore expect DS-FDR to be widely applied in microbiome studies. IMPORTANCE DS-FDR can achieve higher statistical power to detect significant findings in sparse and noisy microbiome data compared to the commonly used Benjamini-Hochberg procedure and other FDR-controlling procedures.

Although genetic approaches are the standard in microbiome analysis, proteome-level information is largely absent. This discrepancy warrants a better understanding of the relationship between gene copy number and protein abundance, as this is crucial information for inferring protein-level changes from metagenomic data. As it remains unknown how metaproteomic systems evolve during dynamic disease states, we leveraged a 4.5-year fecal time series using samples from a single patient with colonic Crohn's disease. Utilizing multiplexed quantitative proteomics and shotgun metagenomic sequencing of eight time points in technical triplicate, we quantified over 29,000 protein groups and 110,000 genes and compared them to five protein biomarkers of disease activity. Broad-scale observations were consistent between data types, including overall clustering by principal-coordinate analysis and fluctuations in Gene Ontology terms related to Crohn's disease. Through linear regression, we determined genes and proteins fluctuating in conjunction with inflammatory metrics. We discovered conserved taxonomic differences relevant to Crohn's disease, including a negative association of Faecalibacterium and a positive association of Escherichia with calprotectin. Despite concordant associations of genera, the specific genes correlated with these metrics were drastically different between metagenomic and metaproteomic data sets. This resulted in the generation of unique functional interpretations dependent on the data type, with metaproteome evidence for previously investigated mechanisms of dysbiosis. An example of one such mechanism was a connection between urease enzymes, amino acid metabolism, and the local inflammation state within the patient. This proof-of-concept approach prompts further investigation of the metaproteome and its relationship with the metagenome in biologically complex systems such as the microbiome. IMPORTANCE A majority of current microbiome research relies heavily on DNA analysis. However, as the field moves toward understanding the microbial functions related to healthy and disease states, it is critical to evaluate how changes in DNA relate to changes in proteins, which are functional units of the genome. This study tracked the abundance of genes and proteins as they fluctuated during various inflammatory states in a 4.5-year study of a patient with colonic Crohn's disease. Our results indicate that despite a low level of correlation, taxonomic associations were consistent in the two data types. While there was overlap of the data types, several associations were uniquely discovered by analyzing the metaproteome component. This case study provides unique and important insights into the fundamental relationship between the genes and proteins of a single individual's fecal microbiome associated with clinical consequences.

Feature selection is indispensable in microbiome data analysis, but it can be particularly challenging as microbiome data sets are high dimensional, underdetermined, sparse and compositional. Great efforts have recently been made on developing new methods for feature selection that handle the above data characteristics, but almost all methods were evaluated based on performance of model predictions. However, little attention has been paid to address a fundamental question: how appropriate are those evaluation criteria? Most feature selection methods often control the model fit, but the ability to identify meaningful subsets of features cannot be evaluated simply based on the prediction accuracy. If tiny changes to the data would lead to large changes in the chosen feature subset, then many selected features are likely to be a data artifact rather than real biological signal. This crucial need of identifying relevant and reproducible features motivated the reproducibility evaluation criterion such as Stability, which quantifies how robust a method is to perturbations in the data. In our paper, we compare the performance of popular model prediction metrics (MSE or AUC) with proposed reproducibility criterion Stability in evaluating four widely used feature selection methods in both simulations and experimental microbiome applications with continuous or binary outcomes. We conclude that Stability is a preferred feature selection criterion over model prediction metrics because it better quantifies the reproducibility of the feature selection method.

Herein, we present a tool called Evident that can be used for deriving effect sizes for a broad spectrum of metadata variables, such as mode of birth, antibiotics, socioeconomics, etc., to provide power calculations for a new study. Evident can be used to mine existing databases of large microbiome studies (such as the American Gut Project, FINRISK, and TEDDY) to analyze the effect sizes for planning future microbiome studies via power analysis. For each metavariable, the Evident software is flexible to compute effect sizes for many commonly used measures of microbiome analyses, including α diversity, β diversity, and log-ratio analysis. In this work, we describe why effect size and power analysis are necessary for computational microbiome analysis and show how Evident can help researchers perform these procedures. Additionally, we describe how Evident is easy for researchers to use and provide an example of efficient analyses using a dataset of thousands of samples and dozens of metadata categories.

Microbiome analyses can be challenging because microbial strains are numerous, and often, confounding factors in the data set are also numerous. Many tools reduce, summarize, and visualize these high-dimensional data to provide insight at the community level. However, they lose the detailed information about each taxon and can be misleading (for example, the well-known horseshoe effect in ordination plots). Thus, multiple methods at different levels of resolution are required to capture the full range of microbial patterns. Here we present Calour, a user-friendly data exploration tool for microbiome analyses. Calour provides a study-centric data model to store and manipulate sample-by-feature tables (with features typically being operational taxonomic units) and their associated metadata. It generates an interactive heatmap, allowing visualization of microbial patterns and exploration using microbial knowledge databases. We demonstrate the use of Calour by exploring publicly available data sets, including the gut and skin microbiota of habitat-switched fire salamander larvae, gut microbiota of Trichuris muris-infected mice, skin microbiota of different human body sites, gut microbiota of various ant species, and a metabolome study of mice exposed to intermittent hypoxia and hypercapnia. In these cases, Calour reveals novel patterns and potential contaminants of subgroups of microbes that are otherwise hard to find. Calour is open source under the Berkeley Software Distribution (BSD) license and available from https://github.com/biocore/calour. IMPORTANCE Calour allows us to identify interesting microbial patterns and generate novel biological hypotheses by interactively inspecting microbiome studies and incorporating annotation databases and convenient statistical tools. Calour can be used as a first-step tool for microbiome data exploration.

Obstructive sleep apnea (OSA), a common sleep disorder characterized by intermittent hypoxia and hypercapnia (IHC), increases atherosclerosis risk. However, the contribution of intermittent hypoxia (IH) or intermittent hypercapnia (IC) in promoting atherosclerosis remains unclear. Since gut microbiota and metabolites have been implicated in atherosclerosis, we examined whether IH or IC alters the microbiome and metabolome to induce a pro-atherosclerotic state. Apolipoprotein E deficient mice (ApoE^-/- ), treated with IH or IC on a high-fat diet (HFD) for 10 weeks, were compared to Air controls. Atherosclerotic lesions were examined, gut microbiome was profiled using 16S rRNA gene amplicon sequencing and metabolome was assessed by untargeted mass spectrometry. In the aorta, IC-induced atherosclerosis was significantly greater than IH and Air controls (aorta, IC 11.1 ± 0.7% vs. IH 7.6 ± 0.4%, p < 0.05 vs. Air 8.1 ± 0.8%, p < 0.05). In the pulmonary artery (PA), however, IH, IC, and Air were significantly different from each other in atherosclerotic formation with the largest lesion observed under IH (PA, IH 40.9 ± 2.0% vs. IC 20.1 ± 2.6% vs. Air 12.2 ± 1.5%, p < 0.05). The most differentially abundant microbial families (p < 0.001) were Peptostreptococcaceae, Ruminococcaceae, and Erysipelotrichaceae. The most differentially abundant metabolites (p < 0.001) were tauro-β-muricholic acid, ursodeoxycholic acid, and lysophosphoethanolamine (18:0). We conclude that IH and IC (a) modulate atherosclerosis progression differently in distinct vascular beds with IC, unlike IH, facilitating atherosclerosis in both aorta and PA and (b) promote an atherosclerotic luminal gut environment that is more evident in IH than IC. We speculate that the resulting changes in the gut metabolome and microbiome interact differently with distinct vascular beds.