Using uncertain data from body-worn sensors to gain insight into type 1 diabetes.

The amount of observational data available for research is growing rapidly with the rise of electronic health records and patient-generated data. However, these data bring new challenges, as data collected outside controlled environments and generated for purposes other than research may be error-prone, biased, or systematically missing. Analysis of these data requires methods that are robust to such challenges, yet methods for causal inference currently only handle uncertainty at the level of causal relationships - rather than variables or specific observations. In contrast, we develop a new approach for causal inference from time series data that allows uncertainty at the level of individual data points, so that inferences depend more strongly on variables and individual observations that are more certain. In the limit, a completely uncertain variable will be treated as if it were not measured. Using simulated data we demonstrate that the approach is more accurate than the state of the art, making substantially fewer false discoveries. Finally, we apply the method to a unique set of data collected from 17 individuals with type 1 diabetes mellitus (T1DM) in free-living conditions over 72h where glucose levels, insulin dosing, physical activity and sleep are measured using body-worn sensors. These data often have high rates of error that vary across time, but we are able to uncover the relationships such as that between anaerobic activity and hyperglycemia. Ultimately, better modeling of uncertainty may enable better translation of methods to free-living conditions, as well as better use of noisy and uncertain EHR data.

A Digital Ecosystem of Diabetes Data and Technology: Services, Systems, and Tools Enabled by Wearables, Sensors, and Apps.

The management of type 1 diabetes (T1D) ideally involves regimented measurement of various health signals; constant interpretation of diverse kinds of data; and consistent cohesion between patients, caregivers, and health care professionals (HCPs). In the context of myriad factors that influence blood glucose dynamics for each individual patient (eg, medication, activity, diet, stress, sleep quality, hormones, environment), such coordination of self-management and clinical care is a great challenge, amplified by the routine unavailability of many types of data thought to be useful in diabetes decision-making. While much remains to be understood about the physiology of diabetes and blood glucose dynamics at the level of the individual, recent and emerging medical and consumer technologies are helping the diabetes community to take great strides toward truly personalized, real-time, data-driven management of this chronic disease. This review describes "connected" technologies--such as smartphone apps, and wearable devices and sensors--which comprise part of a new digital ecosystem of data-driven tools that can link patients and their care teams for precision management of diabetes. These connected technologies are rich sources of physiologic, behavioral, and contextual data that can be integrated and analyzed in "the cloud" for research into personal models of glycemic dynamics, and employed in a multitude of applications for mobile health (mHealth) and telemedicine in diabetes care.

Combining Fourier and lagged k-nearest neighbor imputation for biomedical time series data.

Most clinical and biomedical data contain missing values. A patient's record may be split across multiple institutions, devices may fail, and sensors may not be worn at all times. While these missing values are often ignored, this can lead to bias and error when the data are mined. Further, the data are not simply missing at random. Instead the measurement of a variable such as blood glucose may depend on its prior values as well as that of other variables. These dependencies exist across time as well, but current methods have yet to incorporate these temporal relationships as well as multiple types of missingness. To address this, we propose an imputation method (FLk-NN) that incorporates time lagged correlations both within and across variables by combining two imputation methods, based on an extension to k-NN and the Fourier transform. This enables imputation of missing values even when all data at a time point is missing and when there are different types of missingness both within and across variables. In comparison to other approaches on three biological datasets (simulated and actual Type 1 diabetes datasets, and multi-modality neurological ICU monitoring) the proposed method has the highest imputation accuracy. This was true for up to half the data being missing and when consecutive missing values are a significant fraction of the overall time series length.

Nocturnal continuous glucose and sleep stage data in adults with type 1 diabetes in real-world conditions.

BACKGROUND: Sleep plays an important role in health, and poor sleep is associated with negative impacts on diabetes management, but few studies have objectively evaluated sleep in adults with type 1 diabetes mellitus (T1DM). Nocturnal glycemia and sleep characteristics in T1DM were evaluated using body-worn sensors in real-world conditions. METHODS: Analyses were performed on data collected by the Diabetes Management Integrated Technology Research Initiative pilot study of 17 T1DM subjects: 10 male, 7 female; age 19-61 years; T1DM duration 14.9 ± 11.0 years; hemoglobin A1c (HbA1c) 7.3% ± 1.3% (mean ± standard deviation). Each subject was equipped with a continuous glucose monitor and a wireless sleep monitor (WSM) for four nights. Sleep stages [rapid eye movement (REM), light, and deep sleep] were continuously recorded by the WSM. Nocturnal glycemia (mg/dl) was evaluated as hypoglycemia (<50 mg/dl), low (50-69 mg/dl), euglycemia (70-120 mg/dl), high (121-250 mg/dl), and hyperglycemia (>250 mg/dl) and by several indices of glycemic variability. Glycemia was analyzed within each sleep stage. RESULTS: Subjects slept 358 ± 48 min per night, with 85 ± 27 min in REM sleep, 207 ± 42 min in light sleep, and 66 ± 30 min in deep sleep (mean ± standard deviation). Increased time in deep sleep was associated with lower HbA1c (R2 = 0.42; F = 9.37; p < .01). Nocturnal glycemia varied widely between and within subjects. Glycemia during REM sleep was hypoglycemia 5.5% ± 18.1%, low 6.6% ± 18.5%, euglycemia 44.6% ± 39.5%, high 37.9% ± 39.7%, and hyperglycemia 5.5% ± 21.2%; glycemia during light sleep was hypoglycemia 4.8% ± 12.4%, low 7.3% ± 12.9%, euglycemia 42.1% ± 33.7%, high 39.2% ± 34.6%, and hyperglycemia 6.5% ± 20.5%; and glycemia during deep sleep was hypoglycemia 0.5% ± 2.2%, low 5.8% ± 14.3%, euglycemia 48.0% ± 37.5%, high 39.5% ± 37.6%, and hyperglycemia 6.2% ± 19.5%. Significantly less time was spent in the hypoglycemic range during deep sleep compared with light sleep (p = .02). CONCLUSIONS: Increased time in deep sleep was associated with lower HbA1c, and less hypoglycemia occurred in deep sleep in T1DM, though this must be further evaluated in larger subsequent studies. Furthermore, the consumer-grade WSM device was useful for objectively studying sleep in a real-world setting.

iDASH: integrating data for analysis, anonymization, and sharing.

iDASH (integrating data for analysis, anonymization, and sharing) is the newest National Center for Biomedical Computing funded by the NIH. It focuses on algorithms and tools for sharing data in a privacy-preserving manner. Foundational privacy technology research performed within iDASH is coupled with innovative engineering for collaborative tool development and data-sharing capabilities in a private Health Insurance Portability and Accountability Act (HIPAA)-certified cloud. Driving Biological Projects, which span different biological levels (from molecules to individuals to populations) and focus on various health conditions, help guide research and development within this Center. Furthermore, training and dissemination efforts connect the Center with its stakeholders and educate data owners and data consumers on how to share and use clinical and biological data. Through these various mechanisms, iDASH implements its goal of providing biomedical and behavioral researchers with access to data, software, and a high-performance computing environment, thus enabling them to generate and test new hypotheses.

9p21 DNA variants associated with coronary artery disease impair interferon-γ signalling response.

Genome-wide association studies have identified single nucleotide polymorphisms (SNPs) in the 9p21 gene desert associated with coronary artery disease (CAD) and type 2 diabetes. Despite evidence for a role of the associated interval in neighbouring gene regulation, the biological underpinnings of these genetic associations with CAD or type 2 diabetes have not yet been explained. Here we identify 33 enhancers in 9p21; the interval is the second densest gene desert for predicted enhancers and six times denser than the whole genome (P < 6.55 × 10(-33)). The CAD risk alleles of SNPs rs10811656 and rs10757278 are located in one of these enhancers and disrupt a binding site for STAT1. Lymphoblastoid cell lines homozygous for the CAD risk haplotype show no binding of STAT1, and in lymphoblastoid cell lines homozygous for the CAD non-risk haplotype, binding of STAT1 inhibits CDKN2BAS (also known as CDKN2B-AS1) expression, which is reversed by short interfering RNA knockdown of STAT1. Using a new, open-ended approach to detect long-distance interactions, we find that in human vascular endothelial cells the enhancer interval containing the CAD locus physically interacts with the CDKN2A/B locus, the MTAP gene and an interval downstream of IFNA21. In human vascular endothelial cells, interferon-γ activation strongly affects the structure of the chromatin and the transcriptional regulation in the 9p21 locus, including STAT1-binding, long-range enhancer interactions and altered expression of neighbouring genes. Our findings establish a link between CAD genetic susceptibility and the response to inflammatory signalling in a vascular cell type and thus demonstrate the utility of genome-wide association study findings in directing studies to novel genomic loci and biological processes important for disease aetiology.

Finding distal regulatory elements in the human genome.

Transcriptional regulation of human genes depends not only on promoters and nearby cis-regulatory elements, but also on distal regulatory elements such as enhancers, insulators, locus control regions, and silencing elements, which are often located far away from the genes they control. Our knowledge of human distal regulatory elements is very limited, but the last several years have seen rapid progress in the development of strategies to identify these long-range regulatory sequences throughout the human genome. Here, we review these advances, focusing on two important classes of distal regulatory sequences-enhancers and insulators.

Histone modifications at human enhancers reflect global cell-type-specific gene expression.

The human body is composed of diverse cell types with distinct functions. Although it is known that lineage specification depends on cell-specific gene expression, which in turn is driven by promoters, enhancers, insulators and other cis-regulatory DNA sequences for each gene, the relative roles of these regulatory elements in this process are not clear. We have previously developed a chromatin-immunoprecipitation-based microarray method (ChIP-chip) to locate promoters, enhancers and insulators in the human genome. Here we use the same approach to identify these elements in multiple cell types and investigate their roles in cell-type-specific gene expression. We observed that the chromatin state at promoters and CTCF-binding at insulators is largely invariant across diverse cell types. In contrast, enhancers are marked with highly cell-type-specific histone modification patterns, strongly correlate to cell-type-specific gene expression programs on a global scale, and are functionally active in a cell-type-specific manner. Our results define over 55,000 potential transcriptional enhancers in the human genome, significantly expanding the current catalogue of human enhancers and highlighting the role of these elements in cell-type-specific gene expression.

Distinct and predictive chromatin signatures of transcriptional promoters and enhancers in the human genome.

Eukaryotic gene transcription is accompanied by acetylation and methylation of nucleosomes near promoters, but the locations and roles of histone modifications elsewhere in the genome remain unclear. We determined the chromatin modification states in high resolution along 30 Mb of the human genome and found that active promoters are marked by trimethylation of Lys4 of histone H3 (H3K4), whereas enhancers are marked by monomethylation, but not trimethylation, of H3K4. We developed computational algorithms using these distinct chromatin signatures to identify new regulatory elements, predicting over 200 promoters and 400 enhancers within the 30-Mb region. This approach accurately predicted the location and function of independently identified regulatory elements with high sensitivity and specificity and uncovered a novel functional enhancer for the carnitine transporter SLC22A5 (OCTN2). Our results give insight into the connections between chromatin modifications and transcriptional regulatory activity and provide a new tool for the functional annotation of the human genome.