Bradley E. Aouizerat

Abstract

Background: Accurate prediction of risk for chronic diseases like type 2 diabetes (T2D) is challenging due to the complex underlying etiology. Integration of more complex data types from sensors and leveraging technologies for collection of -omics datasets may provide greater insights into the specific risk profile for complex diseases. Methods: We performed a literature review to identify feature selection methods and machine learning models for prediction of weight loss in a previously completed clinical trial (NCT02278939) of a behavioral intervention for weight loss in Filipinos at risk for T2D. Features included demographic and clinical characteristics, dietary factors, physical activity, and transcriptomics. Results: We identified four feature selection methods: Correlation-based Feature Subset Selection (CfsSubsetEval) with BestFirst, Kolmogorov–Smirnov (KS) test with correlation featureselection (CFS), DESeq2, and max-relevance-min-relevance (MRMR) with linear forward search and mutual information (MI) and four machine learning algorithms: support vector machine, decision tree, random forest, and extra trees that are applicable to prediction of weight loss using the specified feature types. Conclusion: More accurate prediction of risk for T2D and other complex conditions may be possible by leveraging complex data types from sensors and -omics datasets. Emerging methods for feature selection and machine learning algorithms make this type of modeling feasible.

Abstract

Purpose: Gestational diabetes (GDM) is associated with increased risk for preterm birth and related complications for both the pregnant person and newborn. Changes in gene expression have the potential to characterize complex interactions between genetic and behavioral/environmental risk factors for GDM. Our goal was to summarize the state of the science about changes in gene expression and GDM. Design: The systematic review was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Methods: PubMed articles about humans, in English, from any date were included if they described mRNA transcriptome or microRNA findings from blood samples in adults with GDM compared with adults without GDM. Results: Sixteen articles were found representing 1355 adults (n=674 with GDM, n=681 controls) from 12 countries. Three studies reported transcriptome results and thirteen reported microRNA findings. Identified pathways described various aspects of diabetes pathogenesis, including glucose and insulin signaling, regulation, and transport; natural killer cell mediated cytotoxicity; and fatty acid biosynthesis and metabolism. Studies described 135 unique miRNAs that were associated with GDM, of which eight (miR-16-5p, miR-17-5p, miR-20a-5p, miR-29a-3p, miR-195-5p, miR-222-3p, miR-210-3p, and miR-342-3p) were described in 2 or more studies. Findings suggest that miRNA levels vary based on the time in pregnancy when GDM develops, the time point at which they were measured, sex assigned at birth of the offspring, and both the pre-pregnancy and gestational body mass index of the pregnant person. Conclusions: The mRNA, miRNA, gene targets, and pathways identified in this review contribute to our understanding of GDM pathogenesis; however, further research is warranted to validate previous findings. In particular, longitudinal repeated-measures designs are needed that control for participant characteristics (e.g., weight), use standardized data collection methods and analysis tools, and are sufficiently powered to detect differences between subgroups. Findings may be used to improve early diagnosis, prevention, medication choice and/or clinical treatment of patients with GDM.

Abstract

Objectives:This study of people with HIV (PWH) and those without HIV conducted during the COVID-19 pandemic in the United States in 2020 examines the impact of posttraumatic stress disorder (PTSD) on COVID-19 burden, defined as pandemic-related disruptions.Methods:Data consisted of survey responses on PTSD among participants (N = 2434) enrolled in the Multicenter AIDS Cohort Study (MACS) and the Women's Interagency HIV (WIHS) cohorts. Unadjusted and adjusted regression models were used to examine the association of PTSD with COVID-19 burden (overall and domain-specific burdens). Quasi-Poisson regression models were used to assess associations with the COVID-19 burden score and 2 domain-specific burdens: (1) changes in resources and (2) interruptions in health care. Analyses was adjusted for age, race/ethnicity, HIV serostatus, current smoking status, number of comorbidities, education, and study regions.Results:Study participants were a median age of 58 (interquartile range, 52-65) years. In both bivariate and multivariable models, PTSD severity was associated with greater overall COVID-19 burden. PTSD severity was associated with the number of resource changes and number of interruptions in medical care. These findings were also consistent across cohorts (MACS/WIHS) and across HIV serostatus, suggesting a greater risk for COVID-19 burden with greater PTSD severity, which remained significant after controlling for covariates.Conclusions:This study builds on emerging literature demonstrating the impact of mental health on the burden and disruption associated with the COVID-19 pandemic, providing context specific to PWH. The ongoing pandemic requires structural and social interventions to decrease disruption to resources and health resource needs among these vulnerable populations.

Abstract

Neurologic complications of the human immunodeficiency virus (HIV) are common in treated individuals, and toxicity of certain antiretroviral therapies (ART) may contribute to cognitive impairment. We investigated exposures to specific ART and cognition among women living with HIV (WLWH). Virologically suppressed (viral load <200 copies/mL during at least two semi-annual visits) WLWH and age/race matched HIV-seronegative controls enrolled in the Women's Interagency HIV Study who completed at least two biennial cognitive assessments were included. Analysis of WLWH was restricted to those with exposure to the drug class of interest and a nucleoside reverse transcriptase inhibitor (NRTI) backbone. Generalized estimating equations were used to evaluate repeated measures of cognition over time in association with ART class exposure. Among 1,242 eligible WLWH, 20% (n = 247) had isolated drug exposure to non-nucleoside reverse transcriptase inhibitors (NNRTI), 18% (n = 219) to protease inhibitors (PIs), and 6% (n = 79) to integrase inhibitors with a NRTI backbone. Cognitive assessments were performed at a median of 3 biennial visits {IQR 2-4 visits}. At the index assessment, 21% of WLWH demonstrated global cognitive impairment versus 29% at their last cognitive assessment. In multivariable analyses adjusted for hypertension, depression, diabetes mellitus, history of AIDS-defining illness, alcohol use, number of medications, and time on ART, WLWH exposed to NNRTIs demonstrated verbal learning improvements (mean T-score change 1.3, p = .020) compared to other treated women. Compared to HIV-seronegative women, WLWH exposed to PIs had worse verbal learning (mean T-score difference -2.62, p = .002) and verbal memory performance (mean T-score difference -1.74, p = .032) at baseline. Compared to HIV-seronegative women, WLWH exposed to PIs had improvements in verbal learning (mean T-score slope difference 0.36, p = .025) and verbal memory (mean T-score slope difference 0.32, p = .042). The index T-score and slope of change in the T-score were similar among other treated groups and the HIV-seronegative group. We noted emerging trends in cognition in WLWH exposed to specific drug classes. Ongoing study of this relatively young group is important to characterize long-term cognitive outcomes and effect of antiretrovirals as treatment guidelines evolve.

Abstract

Micrornas (mirnas) may be considered impor- tant regulators of risk for type 2 diabetes (T2d). The aim of the present study was to identify novel sets of mirnas associ- ated with T2d risk, as well as their gene and pathway targets. circulating mirnas (n=59) were measured in plasma from participants in a previously completed clinical trial (n=82). an agnostic statistical approach was applied to identify novel sets of mirnas with optimal co-expression patterns. In silico analyses were used to identify the messenger rna and biolog- ical pathway targets of the mirnas within each factor. a total of three factors of miRNAs were identified, containing 18, seven and two mirnas each. eight biological pathways were revealed to contain genes targeted by the mirnas in all three factors, 38 pathways contained genes targeted by the mirnas in two factors, and 55, 18 and two pathways were targeted by the mirnas in a single factor, respectively (all q<0.05). The pathways containing genes targeted by mirnas in the largest factor shared a common theme of biological processes related to metabolism and inflammation. By contrast, the pathways containing genes targeted by mirnas in the second largest factor were related to endocrine function and hormone activity. The present study focused on the pathways uniquely targeted by each factor of mirnas in order to identify unique mecha- nisms that may be associated with a subset of individuals. Further exploration of the genes and pathways related to these biological themes may provide insights about the subtypes of T2D and lead to the identification of novel therapeutic targets.

Abstract

Background: Alterations in gut microbiota and blood metabolomic profiles have been implicated in HIV infection and cardiovascular disease. However, it remains unclear whether alterations in gut microbiota may contribute to disrupted host blood metabolomic profiles in relation to atherosclerosis, especially in the context of HIV infection. Methods: We analyzed cross-sectional associations between gut microbiota features and carotid artery plaque in 361 women with or at high risk of HIV (67% HIV+), and further integrated plaque-associated microbial features with plasma lipidomic/metabolomic profiles. Furthermore, in 737 women and men, we examined prospective associations of baseline gut bacteria-associated lipidomic and metabolomic profiles with incident carotid artery plaque over 7-year follow-up. Results: We found 2 potentially pathogenic bacteria, Fusobacterium and Proteus, were associated with carotid artery plaque; while the beneficial butyrate producer Odoribacter was inversely associated with plaque. Fusobacterium and Proteus were associated with multiple lipids/metabolites which were clustered into 8 modules in network. A module comprised of 9 lysophosphatidylcholines and lysophosphatidylethanolamines and a module comprised of 9 diglycerides were associated with increased risk of carotid artery plaque (risk ratio [95% CI], 1.34 [1.09-1.64] and 1.24 [1.02-1.51] per SD increment, respectively). Functional analyses identified bacterial enzymes in lipid metabolism associated with these plasma lipids. In particular, phospholipase A1 and A2 are the key enzymes in the reactions producing lysophosphatidylcholines and lysophosphatidylethanolamines. Conclusions: Among individuals with or at high risk of HIV infection, we identified altered gut microbiota and related functional capacities in the lipid metabolism associated with disrupted plasma lipidomic profiles and carotid artery atherosclerosis.

Abstract

Here we report three epigenome-wide association studies (EWAS) of DNA methylation on self-reported race, global genetic ancestry, and local genetic ancestry in admixed Americans from three sets of samples, including internal and external replications (Ntotal = 1224). Our EWAS on local ancestry (LA) identified the largest number of ancestry-associated DNA methylation sites and also featured the highest replication rate. Furthermore, by incorporating ancestry origins of genetic variations, we identified 36 methylation quantitative trait loci (meQTL) clumps for LA-associated CpGs that cannot be captured by a model that assumes identical genetic effects across ancestry origins. Lead SNPs at 152 meQTL clumps had significantly different genetic effects in the context of an African or European ancestry background. Local ancestry information enables superior capture of ancestry-associated methylation signatures and identification of ancestry-specific genetic effects on DNA methylation. These findings highlight the importance of incorporating local ancestry for EWAS in admixed samples from multi-ancestry cohorts.

Abstract

Oral squamous cell carcinoma (OSCC) patients suffer from poor survival due to metastasis or locoregional recurrence, processes that are both facilitated by perineural invasion (PNI). OSCC has higher rates of PNI than other cancer subtypes, with PNI present in 80% of tumors. Despite the impact of PNI on oral cancer prognosis and pain, little is known about the genes that drive PNI, which in turn drive pain, invasion, and metastasis. In this study, clinical data, preclinical, and in vitro models are leveraged to elucidate the role of neurotrophins in OSCC metastasis, PNI, and pain. The expression data in OSCC patients with metastasis, PNI, or pain demonstrate dysregulation of neurotrophin genes. TrkA and nerve growth factor receptor (NGFR) are focused, two receptors that are activated by NGF, a neurotrophin expressed at high levels in OSCC. It is demonstrated that targeted knockdown of these two receptors inhibits proliferation and invasion in an in vitro and preclinical model of OSCC, and metastasis, PNI, and pain. It is further determined that TrkA knockdown alone inhibits thermal hyperalgesia, whereas NGFR knockdown alone inhibits mechanical allodynia. Collectively the results highlight the ability of OSCC to co-opt different components of the neurotrophin pathway in metastasis, PNI, and pain.

Abstract

Background: Assessing the effect of alcohol consumption on biological age is essential for understanding alcohol use-related comorbidities and mortality. Previously developed epigenetic clocks are mainly based on DNA methylation in heterogeneous cell types, which provide limited knowledge on the impacts of alcohol consumption at the individual cellular level. Evidence shows that monocytes play an important role in both alcohol-induced pathophysiology and the aging process. In this study, we developed a novel monocyte-based DNA methylation clock (MonoDNAmAge) to assess the impact of alcohol consumption on monocyte age. Methods: A machine learning method was applied to select a set of chronological age-associated DNA methylation CpG sites from 1202 monocyte methylomes. Pearson correlation was tested between MonoDNAmAge and chronological age in three independent cohorts (Ntotal = 2242). Using the MonoDNAmAge clock and four established clocks (i.e., HorvathDNAmAge, HannumDNAmAge, PhenoDNAmAge, GrimDNAmAge), we then evaluated the effect of alcohol consumption on epigenetic aging in the three cohorts [i.e., Yale Stress Center Community Study (YSCCS), Veteran Aging Cohort Study (VACS), Women's Interagency HIV Study (WIHS)] using linear and quadratic models. Results: The MonoDNAmAge, comprised of 186 CpG sites, was moderately to strongly correlated with chronological age in the three cohorts (r = 0.90, p = 3.12E−181 in YSCCS; r = 0.54, p = 1.75E−96 in VACS; r = 0.66, p = 1.50E−60 in WIHS). More importantly, we found a nonlinear association between MonoDNAmAge and alcohol consumption (pmodel = 4.55E−08, (Formula presented.) = 7.80E−08 in YSCCS; pmodel = 1.85E−02, (Formula presented.) = 3.46E−02 in VACS). Heavy alcohol consumption increased EAAMonoDNAmAge up to 1.60 years while light alcohol consumption decreased EAAMonoDNAmAge up to 2.66 years. These results were corroborated by the four established epigenetic clocks (i.e., HorvathDNAmAge, HannumDNAmAge, PhenoDNAmAge, GrimDNAmAge). Conclusions: The results suggest a nonlinear relationship between alcohol consumption and its effects on epigenetic age. Considering adverse effects of alcohol consumption on health, nonlinear effects of alcohol use should be interpreted with caution. The findings, for the first time, highlight the complex effects of alcohol consumption on biological aging.

Abstract

Background: Gene expression is regulated by transcription factors, cofactors, and epigenetic mechanisms. Coexpressed genes indicate similar functional categories and gene networks. Detecting gene-gene coexpression is important for understanding the underlying mechanisms of cellular function and human diseases. A common practice of identifying coexpressed genes is to test the correlation of expression in a set of genes. In single-cell RNA-seq data, an important challenge is the abundance of zero values, so-called “dropout”, which results in biased estimation of gene-gene correlations for downstream analyses. In recent years, efforts have been made to recover coexpressed genes in scRNA-seq data. Here, our goal is to detect coexpressed gene pairs to reduce the “dropout” effect in scRNA-seq data using a novel graph-based k-partitioning method by merging transcriptomically similar cells. Results: We observed that the number of zero values was reduced among the merged transcriptomically similar cell clusters. Motivated by this observation, we leveraged a graph-based algorithm and develop an R package, scCorr, to recover the missing gene-gene correlation in scRNA-seq data that enables the reliable acquisition of cluster-based gene-gene correlations in three independent scRNA-seq datasets. The graphically partitioned cell clusters did not change the local cell community. For example, in scRNA-seq data from peripheral blood mononuclear cells (PBMCs), the gene-gene correlation estimated by scCorr outperformed the correlation estimated by the nonclustering method. Among 85 correlated gene pairs in a set of 100 clusters, scCorr detected 71 gene pairs, while the nonclustering method detected only 4 pairs of a dataset from PBMCs. The performance of scCorr was comparable to those of three previously published methods. As an example of downstream analysis using scCorr, we show that scCorr accurately identified a known cell type (i.e., CD4+ T cells) in PBMCs with a receiver operating characteristic area under the curve of 0.96. Conclusions: Our results demonstrate that scCorr is a robust and reliable graph-based method for identifying correlated gene pairs, which is fundamental to network construction, gene-gene interaction, and cellular omic analyses. scCorr can be quickly and easily implemented to minimize zero values in scRNA-seq analysis and is freely available at https://github.com/CBIIT-CGBB/scCorr.