Identifiability and estimability of Bayesian linear and nonlinear crossed random effects models

Crossed random effects models (CREMs) are particularly useful in longitudinal data applications because they allow researchers to account for the impact of dynamic group membership on individual outcomes. However, no research has determined what data conditions need to be met to sufficiently identify these models, especially the group effects, in a longitudinal context. This is a significant gap in the current literature as future applications to real data may need to consider these conditions to yield accurate and precise model parameter estimates, specifically for the group effects on individual outcomes. Furthermore, there are no existing CREMs that can model intrinsically nonlinear growth. The goals of this study are to develop a Bayesian piecewise CREM to model intrinsically nonlinear growth and evaluate what data conditions are necessary to empirically identify both intrinsically linear and nonlinear longitudinal CREMs. This study includes an applied example that utilizes the piecewise CREM with real data and three simulation studies to assess the data conditions necessary to estimate linear, quadratic, and piecewise CREMs. Results show that the number of repeated measurements collected on groups impacts the ability to recover the group effects. Additionally, functional form complexity impacted data collection requirements for estimating longitudinal CREMs.

Prognostic Performance of Hematological and Serum Iron and Metabolite Indices for Detection of Early Iron Deficiency Induced Metabolic Brain Dysfunction in Infant Rhesus Monkeys

BACKGROUND

The current pediatric practice of monitoring for infantile iron deficiency (ID) via hemoglobin (Hgb) screening at one y of age does not identify preanemic ID nor protect against later neurocognitive deficits.

OBJECTIVES

To identify biomarkers of iron-related metabolic alterations in the serum and brain and determine the sensitivity of conventional iron and heme indices for predicting risk of brain metabolic dysfunction using a nonhuman primate model of infantile ID.

METHODS

Simultaneous serum iron and RBC indices, and serum and cerebrospinal fluid (CSF) metabolomic profiles were determined in 20 rhesus infants, comparing iron sufficient (IS; N = 10) and ID (N = 10) infants at 2 and 4 mo of age.

RESULTS

Reticulocyte hemoglobin (RET-He) was lower at 2 wk in the ID group. Significant IS compared with ID differences in serum iron indices were present at 2 mo, but Hgb and RBC indices differed only at 4 mo (P < 0.05). Serum and CSF metabolomic profiles of the ID and IS groups differed at 2 and 4 mo (P < 0.05). Key metabolites, including homostachydrine and stachydrine (4-5-fold lower at 4 mo in ID group, P < 0.05), were altered in both serum and CSF. Iron indices and RET-He at 2 mo, but not Hgb or other RBC indices, were correlated with altered CSF metabolic profile at 4 mo and had comparable predictive accuracy (area under the receiver operating characteristic curve scores, 0.75-0.80).

CONCLUSIONS

Preanemic ID at 2 mo was associated with metabolic alterations in serum and CSF in infant monkeys. Among the RBC indices, only RET-He predicted the future risk of abnormal CSF metabolic profile with a predictive accuracy comparable to serum iron indices. The concordance of homostachydrine and stachydrine changes in serum and CSF indicates their potential use as early biomarkers of brain metabolic dysfunction in infantile ID.

Multiple augmented reduced rank regression for pan-cancer analysis

Statistical approaches that successfully combine multiple datasets are more powerful, efficient, and scientifically informative than separate analyses. To address variation architectures correctly and comprehensively for high-dimensional data across multiple sample sets (ie, cohorts), we propose multiple augmented reduced rank regression (maRRR), a flexible matrix regression and factorization method to concurrently learn both covariate-driven and auxiliary structured variations. We consider a structured nuclear norm objective that is motivated by random matrix theory, in which the regression or factorization terms may be shared or specific to any number of cohorts. Our framework subsumes several existing methods, such as reduced rank regression and unsupervised multimatrix factorization approaches, and includes a promising novel approach to regression and factorization of a single dataset (aRRR) as a special case. Simulations demonstrate substantial gains in power from combining multiple datasets, and from parsimoniously accounting for all structured variations. We apply maRRR to gene expression data from multiple cancer types (ie, pan-cancer) from The Cancer Genome Atlas, with somatic mutations as covariates. The method performs well with respect to prediction and imputation of held-out data, and provides new insights into mutation-driven and auxiliary variations that are shared or specific to certain cancer types.

Care Management Processes Important for High-Quality Diabetes Care

OBJECTIVE

Identify the improvement in diabetes performance measures and population-based clinical outcomes resulting from changes in care management processes (CMP) in primary care practices over 3 years.

RESEARCH DESIGN AND METHODS

This repeated cross-sectional study tracked clinical performance measures for all diabetes patients seen in a cohort of 330 primary care practices in 2017 and 2019. Unit of analysis was patient-year with practice-level CMP exposures. Causal inference is based on dynamic changes in individual CMPs between years by practice. We used the Bayesian method to simultaneously estimate a five-outcome model: A1c, systolic and diastolic blood pressure, guideline-based statin use, and Optimal Diabetes Care (ODC). We control for unobserved time-invariant practice characteristics and secular change. We modeled correlation of errors across outcomes. Statistical significance was identified using 99% Bayesian credible intervals (analogous to P < 0.01).

RESULTS

Implementation of 18 of 62 CMPs was associated with statistically significant improvements in patient outcomes. Together, these resulted in 12.1% more patients meeting ODC performance measures. Different CMPs affected different outcomes. Three CMPs accounted for 47% of the total ODC improvement, 68% of A1c decrease, 21% of SBP reduction, and 55% of statin use increase: 1) systems for identifying and reminding patients due for testing, 2) after-visit follow-up by a nonclinician, and 3) guideline-based clinician reminders for preventive services during a clinic visit.

CONCLUSIONS

Effective quality improvement in primary care focuses on practice redesign that clearly improves diabetes outcomes. Tailoring CMP adoption in primary care provides effective improvement in ODC performance through focused changes in diabetes outcomes.

Bayesian predictive modeling of multi-source multi-way data

A Bayesian approach to predict a continuous or binary outcome from data that are collected from multiple sources with a multi-way (i.e., multidimensional tensor) structure is described. As a motivating example, molecular data from multiple 'omics sources, each measured over multiple developmental time points, as predictors of early-life iron deficiency (ID) in a rhesus monkey model are considered. The method uses a linear model with a low-rank structure on the coefficients to capture multi-way dependence and model the variance of the coefficients separately across each source to infer their relative contributions. Conjugate priors facilitate an efficient Gibbs sampling algorithm for posterior inference, assuming a continuous outcome with normal errors or a binary outcome with a probit link. Simulations demonstrate that the model performs as expected in terms of misclassification rates and correlation of estimated coefficients with true coefficients, with large gains in performance by incorporating multi-way structure and modest gains when accounting for differing signal sizes across the different sources. Moreover, it provides robust classification of ID monkeys for the motivating application.

Intranasal insulin treatment partially corrects the altered gene expression profile in the hippocampus of developing rats with perinatal iron deficiency

Perinatal iron deficiency (FeD) targets the hippocampus and leads to long-term cognitive deficits. Intranasal insulin administration improves cognitive deficits in adult humans with Alzheimer's disease and type 2 diabetes and could provide benefits in FeD-induced hippocampal dysfunction. To objective was to assess the effects of intranasal insulin administration intranasal insulin administration on the hippocampal transcriptome in a developing rat model of perinatal FeD. Perinatal FeD was induced using low-iron diet from gestational day 3 until postnatal day (P) 7, followed by an iron sufficient (FeS) diet through P21. Intranasal insulin was administered at a dose of 0.3 IU twice daily from P8 to P21. Hippocampi were removed on P21 from FeS control, FeD control, FeS insulin, and FeD insulin groups. Total RNA was isolated and profiled using next-generation sequencing. Gene expression profiles were characterized using custom workflows and expression patterns examined using ingenuity pathways analysis (n = 7-9 per group). Select RNAseq results were confirmed via qPCR. Transcriptomic profiling revealed that mitochondrial biogenesis and flux, oxidative phosphorylation, quantity of neurons, CREB signaling in neurons, and RICTOR-based mTOR signaling were disrupted with FeD and positively affected by intranasal insulin treatment with the most benefit observed in the FeD insulin group. Both perinatal FeD and intranasal insulin administration altered gene expression profile in the developing hippocampus. Intranasal insulin treatment reversed the adverse effects of FeD on many molecular pathways and could be explored as an adjunct therapy in perinatal FeD.

Multiple Augmented Reduced Rank Regression for Pan-Cancer Analysis

Statistical approaches that successfully combine multiple datasets are more powerful, efficient, and scientifically informative than separate analyses. To address variation architectures correctly and comprehensively for high-dimensional data across multiple sample sets (i.e., cohorts), we propose multiple augmented reduced rank regression (maRRR), a flexible matrix regression and factorization method to concurrently learn both covariate-driven and auxiliary structured variation. We consider a structured nuclear norm objective that is motivated by random matrix theory, in which the regression or factorization terms may be shared or specific to any number of cohorts. Our framework subsumes several existing methods, such as reduced rank regression and unsupervised multi-matrix factorization approaches, and includes a promising novel approach to regression and factorization of a single dataset (aRRR) as a special case. Simulations demonstrate substantial gains in power from combining multiple datasets, and from parsimoniously accounting for all structured variation. We apply maRRR to gene expression data from multiple cancer types (i.e., pan-cancer) from TCGA, with somatic mutations as covariates. The method performs well with respect to prediction and imputation of held-out data, and provides new insights into mutation-driven and auxiliary variation that is shared or specific to certain cancer types.

Human Cytomegalovirus in breast milk is associated with milk composition, the infant gut microbiome, and infant growth

Human cytomegalovirus (CMV) is a highly prevalent herpesvirus that is often transmitted to the neonate via breast milk. Postnatal CMV transmission can have negative health consequences for preterm and immunocompromised infants, but any effects on healthy term infants are thought to be benign. Furthermore, the impact of CMV on the composition of the hundreds of bioactive factors in human milk has not been tested. Here, we utilize a cohort of exclusively breastfeeding full term mother-infant pairs to test for differences in the milk transcriptome and metabolome associated with CMV, and the impact of CMV in breast milk on the infant gut microbiome and infant growth. We find upregulation of the indoleamine 2,3- dioxygenase (IDO) tryptophan-to-kynurenine metabolic pathway in CMV+ milk samples, and that CMV+ milk is associated with decreased Bifidobacterium in the infant gut. Our data indicate a complex relationship between milk CMV, milk kynurenine, and infant growth; with kynurenine positively correlated, and CMV viral load negatively correlated, with infant weight-for-length at 1 month of age. These results suggest CMV transmission, CMV-related changes in milk composition, or both may be modulators of full term infant development.

A biomarker discovery framework for childhood anxiety

Introduction

Anxiety is the most common manifestation of psychopathology in youth, negatively affecting academic, social, and adaptive functioning and increasing risk for mental health problems into adulthood. Anxiety disorders are diagnosed only after clinical symptoms emerge, potentially missing opportunities to intervene during critical early prodromal periods. In this study, we used a new empirical approach to extracting nonlinear features of the electroencephalogram (EEG), with the goal of discovering differences in brain electrodynamics that distinguish children with anxiety disorders from healthy children. Additionally, we examined whether this approach could distinguish children with externalizing disorders from healthy children and children with anxiety.

Methods

We used a novel supervised tensor factorization method to extract latent factors from repeated multifrequency nonlinear EEG measures in a longitudinal sample of children assessed in infancy and at ages 3, 5, and 7 years of age. We first examined the validity of this method by showing that calendar age is highly correlated with latent EEG complexity factors (r = 0.77). We then computed latent factors separately for distinguishing children with anxiety disorders from healthy controls using a 5-fold cross validation scheme and similarly for distinguishing children with externalizing disorders from healthy controls.

Results

We found that latent factors derived from EEG recordings at age 7 years were required to distinguish children with an anxiety disorder from healthy controls; recordings from infancy, 3 years, or 5 years alone were insufficient. However, recordings from two (5, 7 years) or three (3, 5, 7 years) recordings gave much better results than 7 year recordings alone. Externalizing disorders could be detected using 3- and 5 years EEG data, also giving better results with two or three recordings than any single snapshot. Further, sex assigned at birth was an important covariate that improved accuracy for both disorder groups, and birthweight as a covariate modestly improved accuracy for externalizing disorders. Recordings from infant EEG did not contribute to the classification accuracy for either anxiety or externalizing disorders.

Conclusion

This study suggests that latent factors extracted from EEG recordings in childhood are promising candidate biomarkers for anxiety and for externalizing disorders if chosen at appropriate ages.

Pneumoproteins and biomarkers of inflammation and coagulation do not predict rapid lung function decline in people living with HIV

Chronic obstructive pulmonary disease (COPD) is among the leading causes of death worldwide and HIV is an independent risk factor for the development of COPD. However, the etiology of this increased risk and means to identify persons with HIV (PWH) at highest risk for COPD have remained elusive. Biomarkers may reveal etiologic pathways and allow better COPD risk stratification. We performed a matched case:control study of PWH in the Strategic Timing of Antiretoviral Treatment (START) pulmonary substudy. Cases had rapid lung function decline (> 40 mL/year FEV₁ decline) and controls had stable lung function (+ 20 to - 20 mL/year). The analysis was performed in two distinct groups: (1) those who were virally suppressed for at least 6 months and (2) those with untreated HIV (from the START deferred treatment arm). We used linear mixed effects models to test the relationship between case:control status and blood concentrations of pneumoproteins (surfactant protein-D and club cell secretory protein), and biomarkers of inflammation (IL-6 and hsCRP) and coagulation (d-dimer and fibrinogen); concentrations were measured within ± 6 months of first included spirometry. We included an interaction with treatment group (untreated HIV vs viral suppression) to test if associations varied by treatment group. This analysis included 77 matched case:control pairs in the virally suppressed batch, and 42 matched case:control pairs in the untreated HIV batch (n = 238 total) who were followed for a median of 3 years. Median (IQR) CD4 + count was lowest in the controls with untreated HIV at 674 (580, 838). We found no significant associations between case:control status and pneumoprotein or biomarker concentrations in either virally suppressed or untreated PWH. In this cohort of relatively young, recently diagnosed PWH, concentrations of pneumoproteins and biomarkers of inflammation and coagulation were not associated with subsequent rapid lung function decline.Trial registration: NCT00867048 and NCT01797367.

Lung proteome and metabolome endotype in HIV-associated obstructive lung disease

Purpose

Obstructive lung disease is increasingly common among persons with HIV, both smokers and nonsmokers. We used aptamer proteomics to identify proteins and associated pathways in HIV-associated obstructive lung disease.

Methods

Bronchoalveolar lavage fluid (BALF) samples from 26 persons living with HIV with obstructive lung disease were matched to persons living with HIV without obstructive lung disease based on age, smoking status and antiretroviral treatment. 6414 proteins were measured using SomaScan® aptamer-based assay. We used sparse distance-weighted discrimination (sDWD) to test for a difference in protein expression and permutation tests to identify univariate associations between proteins and forced expiratory volume in 1 s % predicted (FEV₁ % pred). Significant proteins were entered into a pathway over-representation analysis. We also constructed protein-driven endotypes using K-means clustering and performed over-representation analysis on the proteins that were significantly different between clusters. We compared protein-associated clusters to those obtained from BALF and plasma metabolomics data on the same patient cohort.

Results

After filtering, we retained 3872 proteins for further analysis. Based on sDWD, protein expression was able to separate cases and controls. We found 575 proteins that were significantly correlated with FEV₁ % pred after multiple comparisons adjustment. We identified two protein-driven endotypes, one of which was associated with poor lung function, and found that insulin and apoptosis pathways were differentially represented. We found similar clusters driven by metabolomics in BALF but not plasma.

Conclusion

Protein expression differs in persons living with HIV with and without obstructive lung disease. We were not able to identify specific pathways differentially expressed among patients based on FEV₁ % pred; however, we identified a unique protein endotype associated with insulin and apoptotic pathways.

Human milk variation is shaped by maternal genetics and impacts the infant gut microbiome

Human milk is a complex mix of nutritional and bioactive components that provide complete nutrition for the infant. However, we lack a systematic knowledge of the factors shaping milk composition and how milk variation influences infant health. Here, we used multi-omic profiling to characterize interactions between maternal genetics, milk gene expression, milk composition, and the infant fecal microbiome in 242 exclusively breastfeeding mother-infant pairs. We identified 487 genetic loci associated with milk gene expression unique to the lactating mammary gland, including loci that impacted breast cancer risk and human milk oligosaccharide concentration. Integrative analyses uncovered connections between milk gene expression and infant gut microbiome, including an association between the expression of inflammation-related genes with IL-6 concentration in milk and the abundance of Bifidobacteria in the infant gut. Our results show how an improved understanding of the genetics and genomics of human milk connects lactation biology with maternal and infant health.

Reticulocyte Hemoglobin Equivalent has Comparable Predictive Accuracy as Conventional Serum Iron Indices for Predicting Iron Deficiency and Anemia in a Nonhuman Primate model of Infantile Iron Deficiency

BACKGROUND

Infantile iron deficiency (ID) causes anemia and compromises neurodevelopment. Current screening relies on hemoglobin (Hgb) determination at 1 year of age, which lacks sensitivity and specificity for timely detection of infantile ID. Low reticulocyte Hgb equivalent (RET-He) indicates ID, but its predictive accuracy relative to conventional serum iron indices is unknown.

OBJECTIVES

The objective was to compare diagnostic accuracies of iron indices, red blood cell (RBC) indices, and RET-He for predicting the risk of ID and IDA in a nonhuman primate model of infantile ID.

METHODS

Serum iron, total iron binding capacity, unsaturated iron binding capacity, transferrin saturation (TSAT), Hgb, RET-He, and other RBC indices were determined at 2 wk and 2, 4, and 6 mo in breastfed male and female rhesus infants (N = 54). The diagnostic accuracies of RET-He, iron, and RBC indices for predicting the development of ID (TSAT < 20%) and IDA (Hgb < 10 g/dL + TSAT < 20%) were determined using t tests, area under the receiver operating characteristic curve (AUC) analysis, and multiple regression models.

RESULTS

Twenty-three (42.6%) infants developed ID and 16 (29.6%) progressed to IDA. All 4 iron indices and RET-He, but not Hgb or RBC indices, predicted future risk of ID and IDA (P < 0.001). The predictive accuracy of RET-He (AUC = 0.78, SE = 0.07; P = 0.003) for IDA was comparable to that of the iron indices (AUC = 0.77-0.83, SE = 0.07; P ≤ 0.002). A RET-He threshold of 25.5 pg strongly correlated with TSAT < 20% and correctly predicted IDA in 10 of 16 infants (sensitivity: 62.5%) and falsely predicted possibility of IDA in only 4 of 38 unaffected infants (specificity: 89.5%).

CONCLUSIONS

RET-He is a biomarker of impending ID/IDA in rhesus infants and can be used as a hematological parameter to screen for infantile ID.

Tandem mass tag proteomic and untargeted metabolomic profiling reveals altered serum and CSF biochemical datasets in iron deficient monkeys

The effects of early-life iron deficiency anemia (IDA) extend past the blood and include both short- and long-term adverse effects on many tissues including the brain. Prior to IDA, iron deficiency (ID) can cause similar tissue effects, but a sensitive biomarker of iron-dependent brain health is lacking. To determine serum and CSF biomarkers of ID-induced metabolic dysfunction we performed proteomic and metabolomic analysis of serum and CSF at 4- and 6- months from a nonhuman primate model of infantile IDA. LC/MS/MS analyses identified a total of 227 metabolites and 205 proteins in serum. In CSF, we measured 210 metabolites and 1,560 proteins. Data were either processed from a Q-Exactive (Thermo Scientific, Waltham, MA) through Progenesis QI with accurate mass and retention time comparisons to a proprietary small molecule database and Metlin or with raw files imported directly from a Fusion Orbitrap (Thermo Scientific, Waltham, MA) through Sequest in Proteome Discoverer 2.4.0.305 (Thermo Scientific, Waltham, MA) with peptide matches through the latest Rhesus Macaque HMDB database. Metabolite and protein identifiers, p-values, and q-values were utilized for molecular pathway analysis with Ingenuity Pathways Analysis (IPA). We applied multiway distance weighted discrimination (DWD) to identify a weighted sum of the features (proteins or metabolites) that distinguish ID from IS at 4-months (pre-anemic period) and 6-months of age (anemic).

A natural history study to track brain and spinal cord changes in individuals with Friedreich's ataxia: TRACK-FA study protocol

INTRODUCTION

Drug development for neurodegenerative diseases such as Friedreich's ataxia (FRDA) is limited by a lack of validated, sensitive biomarkers of pharmacodynamic response in affected tissue and disease progression. Studies employing neuroimaging measures to track FRDA have thus far been limited by their small sample sizes and limited follow up. TRACK-FA, a longitudinal, multi-site, and multi-modal neuroimaging natural history study, aims to address these shortcomings by enabling better understanding of underlying pathology and identifying sensitive, clinical trial ready, neuroimaging biomarkers for FRDA.

METHODS

200 individuals with FRDA and 104 control participants will be recruited across seven international study sites. Inclusion criteria for participants with genetically confirmed FRDA involves, age of disease onset ≤ 25 years, Friedreich's Ataxia Rating Scale (FARS) functional staging score of ≤ 5, and a total modified FARS (mFARS) score of ≤ 65 upon enrolment. The control cohort is matched to the FRDA cohort for age, sex, handedness, and years of education. Participants will be evaluated at three study visits over two years. Each visit comprises of a harmonized multimodal Magnetic Resonance Imaging (MRI) and Spectroscopy (MRS) scan of the brain and spinal cord; clinical, cognitive, mood and speech assessments and collection of a blood sample. Primary outcome measures, informed by previous neuroimaging studies, include measures of: spinal cord and brain morphometry, spinal cord and brain microstructure (measured using diffusion MRI), brain iron accumulation (using Quantitative Susceptibility Mapping) and spinal cord biochemistry (using MRS). Secondary and exploratory outcome measures include clinical, cognitive assessments and blood biomarkers.

DISCUSSION

Prioritising immediate areas of need, TRACK-FA aims to deliver a set of sensitive, clinical trial-ready neuroimaging biomarkers to accelerate drug discovery efforts and better understand disease trajectory. Once validated, these potential pharmacodynamic biomarkers can be used to measure the efficacy of new therapeutics in forestalling disease progression.

CLINICAL TRIAL REGISTRATION

ClinicalTrails.gov Identifier: NCT04349514.

Lung and Plasma Metabolome in HIV-Associated Obstructive Lung Disease

BACKGROUND

HIV is a risk factor for obstructive lung disease (OLD), independent of smoking. We used mass spectrometry (MS) approaches to identify metabolomic biomarkers that inform mechanistic pathogenesis of OLD in persons with HIV (PWH).

METHODS

We obtained bronchoalveolar lavage fluid (BALF) samples from 52 PWH, in case:control (+OLD/-OLD) pairs matched on age, smoking status, and antiretroviral treatment. Four hundred nine metabolites from 8 families were measured on BALF and plasma samples using a MS-based Biocrates platform. After filtering metabolites with a high proportion of missing values and values below the level of detection, we performed univariate testing using paired t tests followed by false discovery rate corrections. We used distance-weighted discrimination (DWD) to test for an overall difference in the metabolite profile between cases and controls.

RESULTS

After filtering, there were 252 BALF metabolites for analysis from 8 metabolite families. DWD testing found that collectively, BALF metabolites differentiated cases from controls, whereas plasma metabolites did not. In BALF samples, we identified 3 metabolites that correlated with OLD at the false discovery rate of 10%; all were in the phosphatidylcholine family. We identified additional BALF metabolites when analyzing lung function as a continuous variable, and these included acylcarnitines, triglycerides, and a cholesterol ester.

CONCLUSIONS

Collectively, BALF metabolites differentiate PWH with and without OLD. These included several BALF lipid metabolites. These findings were limited to BALF and were not found in plasma from the same individuals. Phosphatidylcholine, the most common lipid component of surfactant, was the predominant lipid metabolite differentially expressed.

sJIVE: Supervised Joint and Individual Variation Explained

Analyzing multi-source data, which are multiple views of data on the same subjects, has become increasingly common in molecular biomedical research. Recent methods have sought to uncover underlying structure and relationships within and/or between the data sources, and other methods have sought to build a predictive model for an outcome using all sources. However, existing methods that do both are presently limited because they either (1) only consider data structure shared by all datasets while ignoring structures unique to each source, or (2) they extract underlying structures first without consideration to the outcome. The proposed method, supervised joint and individual variation explained (sJIVE), can simultaneously (1) identify shared (joint) and source-specific (individual) underlying structure and (2) build a linear prediction model for an outcome using these structures. These two components are weighted to compromise between explaining variation in the multi-source data and in the outcome. Simulations show sJIVE to outperform existing methods when large amounts of noise are present in the multi-source data. An application to data from the COPDGene study explores gene expression and proteomic patterns associated with lung function.

Multiway sparse distance weighted discrimination

Modern data often take the form of a multiway array. However, most classification methods are designed for vectors, i.e., 1-way arrays. Distance weighted discrimination (DWD) is a popular high-dimensional classification method that has been extended to the multiway context, with dramatic improvements in performance when data have multiway structure. However, the previous implementation of multiway DWD was restricted to classification of matrices, and did not account for sparsity. In this paper, we develop a general framework for multiway classification which is applicable to any number of dimensions and any degree of sparsity. We conducted extensive simulation studies, showing that our model is robust to the degree of sparsity and improves classification accuracy when the data have multiway structure. For our motivating application, magnetic resonance spectroscopy (MRS) was used to measure the abundance of several metabolites across multiple neurological regions and across multiple time points in a mouse model of Friedreich's ataxia, yielding a four-way data array. Our method reveals a robust and interpretable multi-region metabolomic signal that discriminates the groups of interest. We also successfully apply our method to gene expression time course data for multiple sclerosis treatment. An R implementation is available in the package MultiwayClassification at http://github.com/lockEF/MultiwayClassification.

A hierarchical spike-and-slab model for pan-cancer survival using pan-omic data

Background

Pan-omics, pan-cancer analysis has advanced our understanding of the molecular heterogeneity of cancer. However, such analyses have been limited in their ability to use information from multiple sources of data (e.g., omics platforms) and multiple sample sets (e.g., cancer types) to predict clinical outcomes. We address the issue of prediction across multiple high-dimensional sources of data and sample sets by using molecular patterns identified by BIDIFAC+, a method for integrative dimension reduction of bidimensionally-linked matrices, in a Bayesian hierarchical model. Our model performs variable selection through spike-and-slab priors that borrow information across clustered data. We use this model to predict overall patient survival from the Cancer Genome Atlas with data from 29 cancer types and 4 omics sources and use simulations to characterize the performance of the hierarchical spike-and-slab prior.

Results

We found that molecular patterns shared across all or most cancers were largely not predictive of survival. However, our model selected patterns unique to subsets of cancers that differentiate clinical tumor subtypes with markedly different survival outcomes. Some of these subtypes were previously established, such as subtypes of uterine corpus endometrial carcinoma, while others may be novel, such as subtypes within a set of kidney carcinomas. Through simulations, we found that the hierarchical spike-and-slab prior performs best in terms of variable selection accuracy and predictive power when borrowing information is advantageous, but also offers competitive performance when it is not.

Conclusions

We address the issue of prediction across multiple sources of data by using results from BIDIFAC+ in a Bayesian hierarchical model for overall patient survival. By incorporating spike-and-slab priors that borrow information across cancers, we identified molecular patterns that distinguish clinical tumor subtypes within a single cancer and within a group of cancers. We also corroborate the flexibility and performance of using spike-and-slab priors as a Bayesian variable selection approach.

Identification of shared and disease-specific host gene-microbiome associations across human diseases using multi-omic integration

While gut microbiome and host gene regulation independently contribute to gastrointestinal disorders, it is unclear how the two may interact to influence host pathophysiology. Here we developed a machine learning-based framework to jointly analyse paired host transcriptomic (n = 208) and gut microbiome (n = 208) profiles from colonic mucosal samples of patients with colorectal cancer, inflammatory bowel disease and irritable bowel syndrome. We identified associations between gut microbes and host genes that depict shared as well as disease-specific patterns. We found that a common set of host genes and pathways implicated in gastrointestinal inflammation, gut barrier protection and energy metabolism are associated with disease-specific gut microbes. Additionally, we also found that mucosal gut microbes that have been implicated in all three diseases, such as Streptococcus, are associated with different host pathways in each disease, suggesting that similar microbes can affect host pathophysiology in a disease-specific manner through regulation of different host genes. Our framework can be applied to other diseases for the identification of host gene-microbiome associations that may influence disease outcomes.

Multiomic Profiling of Iron Deficient Infant Monkeys Reveals Alterations in Neurologically Important Biochemicals in Serum and CSF Prior to the Onset of Anemia

BACKGROUND

The effects of iron deficiency (ID) during infancy extend beyond the hematologic compartment and include short- and long-term adverse effects on many tissues including the brain. However, sensitive biomarkers of iron-dependent brain health are lacking in humans.

OBJECTIVE

To determine whether serum and CSF biomarkers of ID-induced metabolic dysfunction are concordant in the pre/early anemic stage of ID prior to anemia in a nonhuman primate model of infantile IDA.

METHODS

Paired serum and CSF specimens were collected from iron-sufficient (IS; n = 12) and ID (n = 7) rhesus infants at 4-months (pre-anemic period) and 6-months of age (anemic). Hematological, metabolomic, and proteomic profiles were generated via HPLC/MS at both time points to discriminate serum bio markers of ID-induced brain metabolic dysfunction.

RESULTS

We identified 227 metabolites and 205 proteins in serum. Abnormalities indicating altered liver function, lipid dysregulation, and increased acute phase reactants were present in ID. In CSF, we measured 210 metabolites and 1,560 proteins with changes in ID infants indicative of metabolomic and proteomic differences indexing disrupted synaptogenesis. Systemic and CSF proteomic and metabolomic changes were present and concurrent in the pre-anemic and anemic periods.

CONCLUSIONS

Multiomic serum and CSF profiling uncovered pathways disrupted by ID in both the pre-anemic and anemic stages of infantile IDA, including evidence for hepatic dysfunction and activation of acute phase response. Parallel changes observed in serum and CSF potentially provide measurable serum biomarkers of ID that reflect at-risk brain processes prior to progression to clinical anemia.

BIDIMENSIONAL LINKED MATRIX FACTORIZATION FOR PAN-OMICS PAN-CANCER ANALYSIS

Several modern applications require the integration of multiple large data matrices that have shared rows and/or columns. For example, cancer studies that integrate multiple omics platforms across multiple types of cancer, pan-omics pan-cancer analysis, have extended our knowledge of molecular heterogeneity beyond what was observed in single tumor and single platform studies. However, these studies have been limited by available statistical methodology. We propose a flexible approach to the simultaneous factorization and decomposition of variation across such bidimensionally linked matrices, BIDIFAC+. BIDIFAC+ decomposes variation into a series of low-rank components that may be shared across any number of row sets (e.g., omics platforms) or column sets (e.g., cancer types). This builds on a growing literature for the factorization and decomposition of linked matrices which has primarily focused on multiple matrices that are linked in one dimension (rows or columns) only. Our objective function extends nuclear norm penalization, is motivated by random matrix theory, gives a unique decomposition under relatively mild conditions, and can be shown to give the mode of a Bayesian posterior distribution. We apply BIDIFAC+ to pan-omics pan-cancer data from TCGA, identifying shared and specific modes of variability across four different omics platforms and 29 different cancer types.

Bayesian variable selection in hierarchical difference-in-differences models

A popular method for estimating a causal treatment effect with observational data is the difference-in-differences model. In this work, we consider an extension of the classical difference-in-differences setting to the hierarchical context in which data cannot be matched at the most granular level. Our motivating example is an application to assess the impact of primary care redesign policy on diabetes outcomes in Minnesota, in which the policy is administered at the clinic level and individual outcomes are not matched from pre- to post-intervention. We propose a Bayesian hierarchical difference-in-differences model, which estimates the policy effect by regressing the treatment on a latent variable representing the mean change in group-level outcome. We present theoretical and empirical results showing a hierarchical difference-in-differences model that fails to adjust for a particular class of confounding variables, biases the policy effect estimate. Using a structured Bayesian spike-and-slab model that leverages the temporal structure of the difference-in-differences context, we propose and implement variable selection approaches that target sets of confounding variables leading to unbiased and efficient estimation of the policy effect. We evaluate the methods' properties through simulation, and we use them to assess the impact of primary care redesign of clinics in Minnesota on the management of diabetes outcomes from 2008 to 2017.

Correcting iron deficiency anemia with iron dextran alters the serum metabolomic profile of the infant Rhesus Monkey

BACKGROUND

The effects of infantile iron deficiency anemia (IDA) extend beyond hematological indices and include short- and long-term adverse effects on multiple cells and tissues. IDA is associated with an abnormal serum metabolomic profile, characterized by altered hepatic metabolism, lowered NAD flux, increased nucleoside levels, and a reduction in circulating dopamine levels.

OBJECTIVES

The objective of this study was to determine whether the serum metabolomic profile is normalized after rapid correction of IDA using iron dextran injections.

METHODS

Blood was collected from iron-sufficient (IS; n = 10) and IDA (n = 12) rhesus infants at 6 months of age. IDA infants were then administered iron dextran and vitamin B via intramuscular injections at weekly intervals for 2 to 8 weeks. Their hematological and metabolomic statuses were evaluated following treatment and compared with baseline and a separate group of age-matched IS infants (n = 5).

RESULTS

Serum metabolomic profiles assessed at baseline and after treatment via HPLC/MS using isobaric standards identified 654 quantifiable metabolites. At baseline, 53 metabolites differed between IS and IDA infants. Iron treatment restored traditional hematological indices, including hemoglobin and mean corpuscular volume, into the normal range, but the metabolite profile in the IDA group after iron treatment was markedly altered, with 323 metabolites differentially expressed when compared with an infant's own baseline profile.

CONCLUSIONS

Rapid correction of IDA with iron dextran resulted in extensive metabolic changes across biochemical pathways indexing the liver function, bile acid release, essential fatty acid production, nucleoside release, and several neurologically important metabolites. The results highlight the importance of a cautious approach when developing a route and regimen of iron repletion to treat infantile IDA.

Generalized integrative principal component analysis for multi-type data with block-wise missing structure

High-dimensional multi-source data are encountered in many fields. Despite recent developments on the integrative dimension reduction of such data, most existing methods cannot easily accommodate data of multiple types (e.g. binary or count-valued). Moreover, multi-source data often have block-wise missing structure, i.e. data in one or more sources may be completely unobserved for a sample. The heterogeneous data types and presence of block-wise missing data pose significant challenges to the integration of multi-source data and further statistical analyses. In this article, we develop a low-rank method, called generalized integrative principal component analysis (GIPCA), for the simultaneous dimension reduction and imputation of multi-source block-wise missing data, where different sources may have different data types. We also devise an adapted Bayesian information criterion (BIC) criterion for rank estimation. Comprehensive simulation studies demonstrate the efficacy of the proposed method in terms of rank estimation, signal recovery, and missing data imputation. We apply GIPCA to a mortality study. We achieve accurate block-wise missing data imputation and identify intriguing latent mortality rate patterns with sociological relevance.

Bayesian modeling of associations in bivariate piecewise linear mixed-effects models

Longitudinal processes rarely occur in isolation; often the growth curves of 2 or more variables are interdependent. Moreover, growth curves rarely exhibit a constant pattern of change. Many educational and psychological phenomena are comprised of different developmental phases (segments). Bivariate piecewise linear mixed-effects models (BPLMEM) are a useful and flexible statistical framework that allow simultaneous modeling of 2 processes that portray segmented change and investigates their associations over time. The purpose of the present study was to develop a BPLMEM using a Bayesian inference approach allowing the estimation of the association between the error variances and providing a more robust modeling choice for the joint random-effects of the 2 processes. This study aims to improve upon the limitations of the prior literature on bivariate piecewise mixed-effects models, such as only allowing the modeling of uncorrelated residual errors across the 2 longitudinal processes and restricting modeling choices for the random effects. The performance of the BPLMEM was investigated via a Monte Carlo simulation study. Furthermore, the utility of BPLMEM was illustrated by using a national educational dataset, Early Childhood Longitudinal Study-Kindergarten Cohort (ECLS-K), where we examined the joint development of mathematics and reading achievement scores and the association between their trajectories over 7 measurement occasions. The findings obtained shed new light on the relationship between these 2 prominent educational domains over time. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Early-Life Iron Deficiency and Its Natural Resolution Are Associated with Altered Serum Metabolomic Profiles in Infant Rhesus Monkeys

BACKGROUND

Iron deficiency is the most common nutrient deficiency in human infants aged 6 to 24 mo, and negatively affects many cellular metabolic processes, including energy production, electron transport, and oxidative degradation of toxins. There can be persistent influences on long-term metabolic health beyond its acute effects.

OBJECTIVES

The objective was to determine how iron deficiency in infancy alters the serum metabolomic profile and to test whether these effects persist after the resolution of iron deficiency in a nonhuman primate model of spontaneous iron deficiency.

METHODS

Blood was collected from naturally iron-sufficient (IS; n = 10) and iron-deficient (ID; n = 10) male and female infant rhesus monkeys (Macaca mulatta) at 6 mo of age. Iron deficiency resolved without intervention upon feeding of solid foods, and iron status was re-evaluated at 12 mo of age from the IS and formerly ID monkeys using hematological and other indices; sera were metabolically profiled using HPLC/MS and GC/MS with isobaric standards for identification and quantification at both time points.

RESULTS

A total of 413 metabolites were measured, with differences in 40 metabolites identified between IS and ID monkeys at 6 mo (P$\le $ 0.05). At 12 mo, iron-related hematological parameters had returned to normal, but the formerly ID infants remained metabolically distinct from the age-matched IS infants, with 48 metabolites differentially expressed between the groups. Metabolomic profiling indicated altered liver metabolites, differential fatty acid production, increased serum uridine release, and atypical bile acid production in the ID monkeys.

CONCLUSIONS

Pathway analyses of serum metabolites provided evidence of a hypometabolic state, altered liver function, differential essential fatty acid production, irregular uracil metabolism, and atypical bile acid production in ID infants. Many metabolites remained altered after the resolution of ID, suggesting long-term effects on metabolic health.

A Pan-Cancer and Polygenic Bayesian Hierarchical Model for the Effect of Somatic Mutations on Survival

We built a novel Bayesian hierarchical survival model based on the somatic mutation profile of patients across 50 genes and 27 cancer types. The pan-cancer quality allows for the model to “borrow” information across cancer types, motivated by the assumption that similar mutation profiles may have similar (but not necessarily identical) effects on survival across different tissues of origin or tumor types. The effect of a mutation at each gene was allowed to vary by cancer type, whereas the mean effect of each gene was shared across cancers. Within this framework, we considered 4 parametric survival models (normal, log-normal, exponential, and Weibull), and we compared their performance via a cross-validation approach in which we fit each model on training data and estimate the log-posterior predictive likelihood on test data. The log-normal model gave the best fit, and we investigated the partial effect of each gene on survival via a forward selection procedure. Through this we determined that mutations at TP53 and FAT4 were together the most useful for predicting patient survival. We validated the model via simulation to ensure that our algorithm for posterior computation gave nominal coverage rates. The code used for this analysis can be found at https://github.com/sarahsamorodnitsky/Pan-Cancer-Survival-Modeling.git, and the results are summarized at http://ericfrazerlock.com/surv_figs/SurvivalDisplay.html.

Integrative factorization of bidimensionally linked matrices

Advances in molecular "omics" technologies have motivated new methodologies for the integration of multiple sources of high-content biomedical data. However, most statistical methods for integrating multiple data matrices only consider data shared vertically (one cohort on multiple platforms) or horizontally (different cohorts on a single platform). This is limiting for data that take the form of bidimensionally linked matrices (eg, multiple cohorts measured on multiple platforms), which are increasingly common in large-scale biomedical studies. In this paper, we propose bidimensional integrative factorization (BIDIFAC) for integrative dimension reduction and signal approximation of bidimensionally linked data matrices. Our method factorizes data into (a) globally shared, (b) row-shared, (c) column-shared, and (d) single-matrix structural components, facilitating the investigation of shared and unique patterns of variability. For estimation, we use a penalized objective function that extends the nuclear norm penalization for a single matrix. As an alternative to the complicated rank selection problem, we use results from the random matrix theory to choose tuning parameters. We apply our method to integrate two genomics platforms (messenger RNA and microRNA expression) across two sample cohorts (tumor samples and normal tissue samples) using the breast cancer data from the Cancer Genome Atlas. We provide R code for fitting BIDIFAC, imputing missing values, and generating simulated data.

Bayesian GWAS with Structured and Non-Local Priors

MOTIVATION

The flexibility of a Bayesian framework is promising for GWAS, but current approaches can benefit from more informative prior models. We introduce a novel Bayesian approach to GWAS, called Structured and Non-Local Priors (SNLPs) GWAS, that improves over existing methods in two important ways. First, we describe a model that allows for a marker's gene-parent membership and other characteristics to influence its probability of association with an outcome. Second, we describe a non-local alternative model for differential minor allele rates at each marker, in which the null and alternative hypotheses have no common support.

RESULTS

We employ a non-parametric model that allows for clustering of the genes in tandem with a regression model for marker-level covariates, and demonstrate how incorporating these additional characteristics can improve power. We further demonstrate that our non-local alternative model gives symmetric rates of convergence for the null and alternative hypotheses, whereas commonly used local alternative models have asymptotic rates that favor the alternative hypothesis over the null. We demonstrate the robustness and flexibility of our structured and non-local model for different data generating scenarios and signal-to-noise ratios. We apply our Bayesian GWAS method to single nucleotide polymorphisms data collected from a pool of Alzheimer's disease and cognitively normal patients from the Alzheimer's Database Neuroimaging Initiative.

AVAILABILITY AND IMPLEMENTATION

R code to perform the SNLPs method is available at https://github.com/lockEF/BayesianScreening.

Linked matrix factorization

Several recent methods address the dimension reduction and decomposition of linked high-content data matrices. Typically, these methods consider one dimension, rows or columns, that is shared among the matrices. This shared dimension may represent common features measured for different sample sets (horizontal integration) or a common sample set with features from different platforms (vertical integration). We introduce an approach for simultaneous horizontal and vertical integration, Linked Matrix Factorization (LMF), for the general case where some matrices share rows (e.g., features) and some share columns (e.g., samples). Our motivating application is a cytotoxicity study with accompanying genomic and molecular chemical attribute data. The toxicity matrix (cell lines × chemicals) shares samples with a genotype matrix (cell lines × SNPs) and shares features with a molecular attribute matrix (chemicals × attributes). LMF gives a unified low-rank factorization of these three matrices, which allows for the decomposition of systematic variation that is shared and systematic variation that is specific to each matrix. This allows for efficient dimension reduction, exploratory visualization, and the imputation of missing data even when entire rows or columns are missing. We present theoretical results concerning the uniqueness, identifiability, and minimal parametrization of LMF, and evaluate it with extensive simulation studies.

Detecting Multiple Random Changepoints in Bayesian Piecewise Growth Mixture Models

Piecewise growth mixture models are a flexible and useful class of methods for analyzing segmented trends in individual growth trajectory over time, where the individuals come from a mixture of two or more latent classes. These models allow each segment of the overall developmental process within each class to have a different functional form; examples include two linear phases of growth, or a quadratic phase followed by a linear phase. The changepoint (knot) is the time of transition from one developmental phase (segment) to another. Inferring the location of the changepoint(s) is often of practical interest, along with inference for other model parameters. A random changepoint allows for individual differences in the transition time within each class. The primary objectives of our study are as follows: (1) to develop a PGMM using a Bayesian inference approach that allows the estimation of multiple random changepoints within each class; (2) to develop a procedure to empirically detect the number of random changepoints within each class; and (3) to empirically investigate the bias and precision of the estimation of the model parameters, including the random changepoints, via a simulation study. We have developed the user-friendly package BayesianPGMM for R to facilitate the adoption of this methodology in practice, which is available at https://github.com/lockEF/BayesianPGMM . We describe an application to mouse-tracking data for a visual recognition task.

Multi-omic molecular profiling of lung cancer in COPD

Chronic obstructive pulmonary disease (COPD) is a known risk factor for developing lung cancer but the underlying mechanisms remain unknown. We hypothesise that the COPD stroma contains molecular mechanisms supporting tumourigenesis.We conducted an unbiased multi-omic analysis to identify gene expression patterns that distinguish COPD stroma in patients with or without lung cancer. We obtained lung tissue from patients with COPD and lung cancer (tumour and adjacent non-malignant tissue) and those with COPD without lung cancer for profiling of proteomic and mRNA (both cytoplasmic and polyribosomal). We used the Joint and Individual Variation Explained (JIVE) method to integrate and analyse across the three datasets.JIVE identified eight latent patterns that robustly distinguished and separated the three groups of tissue samples (tumour, adjacent and control). Predictive variables that associated with the tumour, compared to adjacent stroma, were mainly represented in the transcriptomic data, whereas predictive variables associated with adjacent tissue, compared to controls, were represented at the translatomic level. Pathway analysis revealed extracellular matrix and phosphatidylinositol-4,5-bisphosphate 3-kinase-protein kinase B signalling pathways as important signals in the tumour adjacent stroma.The multi-omic approach distinguishes tumour adjacent stroma in lung cancer and reveals two stromal expression patterns associated with cancer.

Tensor-on-tensor regression

We propose a framework for the linear prediction of a multi-way array (i.e., a tensor) from another multi-way array of arbitrary dimension, using the contracted tensor product. This framework generalizes several existing approaches, including methods to predict a scalar outcome from a tensor, a matrix from a matrix, or a tensor from a scalar. We describe an approach that exploits the multiway structure of both the predictors and the outcomes by restricting the coefficients to have reduced CP-rank. We propose a general and efficient algorithm for penalized least-squares estimation, which allows for a ridge (L 2) penalty on the coefficients. The objective is shown to give the mode of a Bayesian posterior, which motivates a Gibbs sampling algorithm for inference. We illustrate the approach with an application to facial image data. An R package is available at https://github.com/lockEF/MultiwayRegression.

Supervised multiway factorization

We describe a probabilistic PARAFAC/CANDECOMP (CP) factorization for multiway (i.e., tensor) data that incorporates auxiliary covariates, SupCP. SupCP generalizes the supervised singular value decomposition (SupSVD) for vector-valued observations, to allow for observations that have the form of a matrix or higher-order array. Such data are increasingly encountered in biomedical research and other fields. We use a novel likelihood-based latent variable representation of the CP factorization, in which the latent variables are informed by additional covariates. We give conditions for identifiability, and develop an EM algorithm for simultaneous estimation of all model parameters. SupCP can be used for dimension reduction, capturing latent structures that are more accurate and interpretable due to covariate supervision. Moreover, SupCP specifies a full probability distribution for a multiway data observation with given covariate values, which can be used for predictive modeling. We conduct comprehensive simulations to evaluate the SupCP algorithm. We apply it to a facial image database with facial descriptors (e.g., smiling / not smiling) as covariates, and to a study of amino acid fluorescence. Software is available at https://github.com/lockEF/SupCP.

Prediction With Dimension Reduction of Multiple Molecular Data Sources for Patient Survival

Predictive modeling from high-dimensional genomic data is often preceded by a dimension reduction step, such as principal component analysis (PCA). However, the application of PCA is not straightforward for multisource data, wherein multiple sources of ‘omics data measure different but related biological components. In this article, we use recent advances in the dimension reduction of multisource data for predictive modeling. In particular, we apply exploratory results from Joint and Individual Variation Explained (JIVE), an extension of PCA for multisource data, for prediction of differing response types. We conduct illustrative simulations to illustrate the practical advantages and interpretability of our approach. As an application example, we consider predicting survival for patients with glioblastoma multiforme from 3 data sources measuring messenger RNA expression, microRNA expression, and DNA methylation. We also introduce a method to estimate JIVE scores for new samples that were not used in the initial dimension reduction and study its theoretical properties; this method is implemented in the R package R.JIVE on CRAN, in the function jive.predict.

Discriminating sample groups with multi-way data

High-dimensional linear classifiers, such as distance weighted discrimination (DWD) and versions of the support vector machine (SVM), are commonly used in biomedical research to distinguish groups of subjects based on a large number of features. However, their use is limited to applications where a single vector of features is measured for each subject. In practice, data are often multi-way, or measured over multiple dimensions. For example, metabolite abundance may be measured over multiple regions or tissues, or gene expression may be measured over multiple time points, for the same subjects. We propose a framework for linear classification of high-dimensional multi-way data, in which coefficients can be factorized into weights that are specific to each dimension. More generally, the coefficients for each measurement in a multi-way dataset are assumed to have low-rank structure. This framework extends existing classification techniques from single vector to multi-way features, and we have implemented multi-way versions of SVM and DWD. We describe informative simulation results, and apply multi-way DWD to data for two very different clinical research studies. The first study uses magnetic resonance spectroscopy metabolite data over multiple brain regions to compare participants with and without spinocerebellar ataxia; the second uses publicly available gene expression time-course data to compare degrees of treatment response among patients with multiple sclerosis. Our multi-way method can improve performance and simplify interpretation over naive applications of full rank linear and non-linear classification to multi-way data. The R package is available at https://github.com/lockEF/MultiwayClassification.

Bayesian genome- and epigenome-wide association studies with gene level dependence

High-throughput genetic and epigenetic data are often screened for associations with an observed phenotype. For example, one may wish to test hundreds of thousands of genetic variants, or DNA methylation sites, for an association with disease status. These genomic variables can naturally be grouped by the gene they encode, among other criteria. However, standard practice in such applications is independent screening with a universal correction for multiplicity. We propose a Bayesian approach in which the prior probability of an association for a given genomic variable depends on its gene, and the gene-specific probabilities are modeled nonparametrically. This hierarchical model allows for appropriate gene and genome-wide multiplicity adjustments, and can be incorporated into a variety of Bayesian association screening methodologies with negligible increase in computational complexity. We describe an application to screening for differences in DNA methylation between lower grade glioma and glioblastoma multiforme tumor samples from The Cancer Genome Atlas. Software is available via the package BayesianScreening for R: github.com/lockEF/BayesianScreening.

Metabolomic analysis of CSF indicates brain metabolic impairment precedes hematological indices of anemia in the iron-deficient infant monkey

OBJECTIVES

Iron deficiency (ID) anemia leads to long-term neurodevelopmental deficits by altering iron-dependent brain metabolism. The objective of the study was to determine if ID induces metabolomic abnormalities in the cerebrospinal fluid (CSF) in the pre-anemic stage and to ascertain the aspects of abnormal brain metabolism affected.

METHODS

Standard hematological parameters [hemoglobin (Hgb), mean corpuscular volume (MCV), transferrin (Tf) saturation, and zinc protoporphyrin/heme (ZnPP/H)] were compared at 2, 4, 6, 8, and 12 months in iron-sufficient (IS; n = 7) and iron-deficient (ID; n = 7) infant rhesus monkeys. Five CSF metabolite ratios were determined at 4, 8, and 12 months using ¹H NMR spectroscopy at 16.4 T and compared between groups and in relation to hematologic parameters.

RESULTS

ID infants developed ID (Tf saturation < 25%) by 4 months of age and all became anemic (Hgb < 110 g/L and MCV < 60 fL) at 6 months. Their heme indices normalized by 12 months. Pyruvate/glutamine and phosphocreatine/creatine (PCr/Cr) ratios in CSF were lower in the ID infants by 4 months (P < 0.05). The PCr/Cr ratio remained lower at 8 months (P = 0.02). ZnPP/H, an established blood marker of pre-anemic ID, was positively correlated with the CSF citrate/glutamine ratio (marginal correlation, 0.34; P < 0.001; family wise error rate = 0.001).

DISCUSSION

Metabolomic analysis of the CSF is sensitive for detecting the effects of pre-anemic ID on brain energy metabolism. Persistence of a lower PCr/Cr ratio at 8 months, even as hematological measures demonstrated recovery from anemia, indicate that the restoration of brain energy metabolism is delayed. Metabolomic platforms offer a useful tool for early detection of the impact of ID on brain metabolism in infants.

R.JIVE for exploration of multi-source molecular data

: The integrative analysis of multiple high-throughput data sources that are available for a common sample set is an increasingly common goal in biomedical research. Joint and individual variation explained (JIVE) is a tool for exploratory dimension reduction that decomposes a multi-source dataset into three terms: a low-rank approximation capturing joint variation across sources, low-rank approximations for structured variation individual to each source and residual noise. JIVE has been used to explore multi-source data for a variety of application areas but its accessibility was previously limited. We introduce R.JIVE, an intuitive R package to perform JIVE and visualize the results. We discuss several improvements and extensions of the JIVE methodology that are included. We illustrate the package with an application to multi-source breast tumor data from The Cancer Genome Atlas.

AVAILABILITY AND IMPLEMENTATION

R.JIVE is available via the Comprehensive R Archive Network (CRAN) under the GPLv3 license: https://cran.r-project.org/web/packages/r.jive/

CONTACT

elock@umn.edu

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Analysis of multi-source metabolomic data using joint and individual variation explained (JIVE)

Metabolic profiling is increasingly being used for understanding biological processes but there is no single analytical technique that provides a complete quantitative or qualitative profiling of the metabolome. Data fusion (i.e. joint analysis of data from multiple sources) has the potential to circumvent this issue facilitating knowledge discovery and reliable biomarker identification. Another field of application of data fusion is the simultaneous analysis of metabolomic changes through several biofluids or tissues. However, metabolomics typically deals with large datasets, with hundreds to thousands of variables and the identification of shared and individual factors or structures across multiple sources is challenging due to the high variable to sample ratios and differences in intensity and noise range. In this work we apply a recent method, Joint and Individual Variation Explained (JIVE), for the integrated unsupervised analysis of metabolomic profiles from multiple data sources. This method separates the shared patterns among data sources (i.e. the joint structure) from the individual structure of each data source that is unrelated to the joint structure. Two examples are described to show the applicability of JIVE for the simultaneous analysis of multi-source data using: (i) plasma samples subjected to different analytical techniques, sample treatment and measurement conditions; and (ii) plasma and urine samples subjected to liquid chromatography-mass spectrometry measured using two ionization conditions.

Shared kernel Bayesian screening

This article concerns testing for equality of distribution between groups. We focus on screening variables with shared distributional features such as common support, modes and patterns of skewness. We propose a Bayesian testing method using kernel mixtures, which improves performance by borrowing information across the different variables and groups through shared kernels and a common probability of group differences. The inclusion of shared kernels in a finite mixture, with Dirichlet priors on the weights, leads to a simple framework for testing that scales well for high-dimensional data. We provide closed asymptotic forms for the posterior probability of equivalence in two groups and prove consistency under model misspecification. The method is applied to DNA methylation array data from a breast cancer study, and compares favourably to competitors when Type I error is estimated via permutation.

Joint eQTL assessment of whole blood and dura mater tissue from individuals with Chiari type I malformation

BACKGROUND

Expression quantitative trait loci (eQTL) play an important role in the regulation of gene expression. Gene expression levels and eQTLs are expected to vary from tissue to tissue, and therefore multi-tissue analyses are necessary to fully understand complex genetic conditions in humans. Dura mater tissue likely interacts with cranial bone growth and thus may play a role in the etiology of Chiari Type I Malformation (CMI) and related conditions, but it is often inaccessible and its gene expression has not been well studied. A genetic basis to CMI has been established; however, the specific genetic risk factors are not well characterized.

RESULTS

We present an assessment of eQTLs for whole blood and dura mater tissue from individuals with CMI. A joint-tissue analysis identified 239 eQTLs in either dura or blood, with 79% of these eQTLs shared by both tissues. Several identified eQTLs were novel and these implicate genes involved in bone development (IPO8, XYLT1, and PRKAR1A), and ribosomal pathways related to marrow and bone dysfunction, as potential candidates in the development of CMI.

CONCLUSIONS

Despite strong overall heterogeneity in expression levels between blood and dura, the majority of cis-eQTLs are shared by both tissues. The power to detect shared eQTLs was improved by using an integrative statistical approach. The identified tissue-specific and shared eQTLs provide new insight into the genetic basis for CMI and related conditions.

Bayesian consensus clustering

MOTIVATION

In biomedical research a growing number of platforms and technologies are used to measure diverse but related information, and the task of clustering a set of objects based on multiple sources of data arises in several applications. Most current approaches to multisource clustering either independently determine a separate clustering for each data source or determine a single 'joint' clustering for all data sources. There is a need for more flexible approaches that simultaneously model the dependence and the heterogeneity of the data sources.

RESULTS

We propose an integrative statistical model that permits a separate clustering of the objects for each data source. These separate clusterings adhere loosely to an overall consensus clustering, and hence they are not independent. We describe a computationally scalable Bayesian framework for simultaneous estimation of both the consensus clustering and the source-specific clusterings. We demonstrate that this flexible approach is more robust than joint clustering of all data sources, and is more powerful than clustering each data source independently. We present an application to subtype identification of breast cancer tumor samples using publicly available data from The Cancer Genome Atlas.

AVAILABILITY

R code with instructions and examples is available at http://people.duke.edu/%7Eel113/software.html.

JOINT AND INDIVIDUAL VARIATION EXPLAINED (JIVE) FOR INTEGRATED ANALYSIS OF MULTIPLE DATA TYPES

Research in several fields now requires the analysis of datasets in which multiple high-dimensional types of data are available for a common set of objects. In particular, The Cancer Genome Atlas (TCGA) includes data from several diverse genomic technologies on the same cancerous tumor samples. In this paper we introduce Joint and Individual Variation Explained (JIVE), a general decomposition of variation for the integrated analysis of such datasets. The decomposition consists of three terms: a low-rank approximation capturing joint variation across data types, low-rank approximations for structured variation individual to each data type, and residual noise. JIVE quantifies the amount of joint variation between data types, reduces the dimensionality of the data, and provides new directions for the visual exploration of joint and individual structure. The proposed method represents an extension of Principal Component Analysis and has clear advantages over popular two-block methods such as Canonical Correlation Analysis and Partial Least Squares. A JIVE analysis of gene expression and miRNA data on Glioblastoma Multiforme tumor samples reveals gene-miRNA associations and provides better characterization of tumor types.

ToxPi GUI: an interactive visualization tool for transparent integration of data from diverse sources of evidence

MOTIVATION

Scientists and regulators are often faced with complex decisions, where use of scarce resources must be prioritized using collections of diverse information. The Toxicological Prioritization Index (ToxPi™) was developed to enable integration of multiple sources of evidence on exposure and/or safety, transformed into transparent visual rankings to facilitate decision making. The rankings and associated graphical profiles can be used to prioritize resources in various decision contexts, such as testing chemical toxicity or assessing similarity of predicted compound bioactivity profiles. The amount and types of information available to decision makers are increasing exponentially, while the complex decisions must rely on specialized domain knowledge across multiple criteria of varying importance. Thus, the ToxPi bridges a gap, combining rigorous aggregation of evidence with ease of communication to stakeholders.

RESULTS

An interactive ToxPi graphical user interface (GUI) application has been implemented to allow straightforward decision support across a variety of decision-making contexts in environmental health. The GUI allows users to easily import and recombine data, then analyze, visualize, highlight, export and communicate ToxPi results. It also provides a statistical metric of stability for both individual ToxPi scores and relative prioritized ranks.

AVAILABILITY

The ToxPi GUI application, complete user manual and example data files are freely available from http://comptox.unc.edu/toxpi.php.

Quantitative high-throughput screening for chemical toxicity in a population-based in vitro model

A shift in toxicity testing from in vivo to in vitro may efficiently prioritize compounds, reveal new mechanisms, and enable predictive modeling. Quantitative high-throughput screening (qHTS) is a major source of data for computational toxicology, and our goal in this study was to aid in the development of predictive in vitro models of chemical-induced toxicity, anchored on interindividual genetic variability. Eighty-one human lymphoblast cell lines from 27 Centre d'Etude du Polymorphisme Humain trios were exposed to 240 chemical substances (12 concentrations, 0.26nM-46.0μM) and evaluated for cytotoxicity and apoptosis. qHTS screening in the genetically defined population produced robust and reproducible results, which allowed for cross-compound, cross-assay, and cross-individual comparisons. Some compounds were cytotoxic to all cell types at similar concentrations, whereas others exhibited interindividual differences in cytotoxicity. Specifically, the qHTS in a population-based human in vitro model system has several unique aspects that are of utility for toxicity testing, chemical prioritization, and high-throughput risk assessment. First, standardized and high-quality concentration-response profiling, with reproducibility confirmed by comparison with previous experiments, enables prioritization of chemicals for variability in interindividual range in cytotoxicity. Second, genome-wide association analysis of cytotoxicity phenotypes allows exploration of the potential genetic determinants of interindividual variability in toxicity. Furthermore, highly significant associations identified through the analysis of population-level correlations between basal gene expression variability and chemical-induced toxicity suggest plausible mode of action hypotheses for follow-up analyses. We conclude that as the improved resolution of genetic profiling can now be matched with high-quality in vitro screening data, the evaluation of the toxicity pathways and the effects of genetic diversity are now feasible through the use of human lymphoblast cell lines.

Interstrain differences in the liver effects of trichloroethylene in a multistrain panel of inbred mice

Trichloroethylene (TCE) is a widely used industrial chemical and a common environmental contaminant. It is a well-known carcinogen in rodents and a probable carcinogen in humans. Studies utilizing panels of mouse inbred strains afford a unique opportunity to understand both metabolic and genetic basis for differences in responses to TCE. We tested the hypothesis that strain- and liver-specific toxic effects of TCE are genetically controlled and that the mechanisms of toxicity and susceptibility can be uncovered by exploring responses to TCE using a diverse panel of inbred mouse strains. TCE (2100 mg/kg) or corn oil vehicle was administered by gavage to 6- to 8-week-old male mice of 15 mouse strains. Serum and liver were collected at 2, 8, and 24 h postdosing and were analyzed for TCE metabolites, hepatocellular injury, and gene expression of liver. TCE metabolism, as evident from the levels of individual oxidative and conjugative metabolites, varied considerably between strains. TCE treatment-specific effect on the liver transcriptome was strongly dependent on genetic background. Peroxisome proliferator-activated receptor-mediated molecular networks, consisting of the metabolism genes known to be induced by TCE, represent some of the most pronounced molecular effects of TCE treatment in mouse liver that are dependent on genetic background. Conversely, cell death, liver necrosis, and immune-mediated response pathways, which are altered by TCE treatment in liver, are largely genetic background independent. These studies provide better understanding of the mechanisms of TCE-induced toxicity anchored on metabolism and genotype-phenotype correlations that may define susceptibility or resistance.