Elastic Net Model Research Articles

Radiomic Modeling of Bone Density and Rib Fracture Risk After Stereotactic Body Radiation Therapy for Early-Stage Non-Small Cell Lung Cancer

PurposeOur purpose was to determine whether bone density and bone-derived radiomic metrics in combination with dosimetric variables could improve risk stratification of rib fractures after stereotactic body radiation therapy (SBRT) for early-stage non-small cell lung cancer (NSCLC). Methods and MaterialsA retrospective analysis was conducted of patients with early-stage NSCLC treated with SBRT. Dosimetric data and rib radiomic data extracted using PyRadiomics were used for the analysis. A subset of patients had bone density scans that were used to create a predicted bone density score for all patients. A 10-fold cross validated approach with 10 resamples was used to find the top univariate logistic models and elastic net regression models that predicted for rib fracture. ResultsA total of 192 treatment plans were included in the study with a rib fracture rate of 16.1%. A predicted bone density score was created from a multivariate model with vertebral body Hounsfield units and patient weight, with an R-squared of 0.518 compared with patient dual-energy x-ray absorptiometry T-scores. When analyzing all patients, a low predicted bone density score approached significance for increased risk of rib fracture (P = .07). On competing risk analysis, when stratifying patients based on chest wall V30 Gy and bone density score, those with a V30 Gy ≥30 cc and a low bone density score had a significantly higher risk of rib fracture compared with all other patients (P < .001), with a predicted 2-year risk of rib fracture of 28.6% (95% confidence interval, 17.2%-41.1%) and 4.9% (95% confidence interval, 2.3%-9.0%), respectively. Dosimetric variables were the primary drivers of fracture risk. A multivariate elastic net regression model including all dosimetric variables was the best predictor of rib fracture (area under the curve [AUC], 0.864). Bone density variables (AUC, 0.618) and radiomic variables (AUC, 0.617) have better predictive power than clinical variables that exclude bone density (AUC, 0.538). ConclusionRadiomic features, including a bone density score that includes vertebral body Hounsfield units and radiomic signatures from the ribs, can be used to stratify risk of rib fracture after SBRT for NSCLC.

Capturing SNP Association across the NK Receptor and HLA Gene Regions in Multiple Sclerosis by Targeted Penalised Regression Models.

Conventional genome-wide association studies (GWASs) of complex traits, such as Multiple Sclerosis (MS), are reliant on per-SNP p-values and are therefore heavily burdened by multiple testing correction. Thus, in order to detect more subtle alterations, ever increasing sample sizes are required, while ignoring potentially valuable information that is readily available in existing datasets. To overcome this, we used penalised regression incorporating elastic net with a stability selection method by iterative subsampling to detect the potential interaction of loci with MS risk. Through re-analysis of the ANZgene dataset (1617 cases and 1988 controls) and an IMSGC dataset as a replication cohort (1313 cases and 1458 controls), we identified new association signals for MS predisposition, including SNPs above and below conventional significance thresholds while targeting two natural killer receptor loci and the well-established HLA loci. For example, rs2844482 (98.1% iterations), otherwise ignored by conventional statistics (p = 0.673) in the same dataset, was independently strongly associated with MS in another GWAS that required more than 40 times the number of cases (~45 K). Further comparison of our hits to those present in a large-scale meta-analysis, confirmed that the majority of SNPs identified by the elastic net model reached conventional statistical GWAS thresholds (p < 5 × 10−8) in this much larger dataset. Moreover, we found that gene variants involved in oxidative stress, in addition to innate immunity, were associated with MS. Overall, this study highlights the benefit of using more advanced statistical methods to (re-)analyse subtle genetic variation among loci that have a biological basis for their contribution to disease risk.

Predicting Fall Counts Using Wearable Sensors: A Novel Digital Biomarker for Parkinson's Disease.

People with Parkinson’s disease (PD) experience significant impairments to gait and balance; as a result, the rate of falls in people with Parkinson’s disease is much greater than that of the general population. Falls can have a catastrophic impact on quality of life, often resulting in serious injury and even death. The number (or rate) of falls is often used as a primary outcome in clinical trials on PD. However, falls data can be unreliable, expensive and time-consuming to collect. We sought to validate and test a novel digital biomarker for PD that uses wearable sensor data obtained during the Timed Up and Go (TUG) test to predict the number of falls that will be experienced by a person with PD. Three datasets, containing a total of 1057 (671 female) participants, including 71 previously diagnosed with PD, were included in the analysis. Two statistical approaches were considered in predicting falls counts: the first based on a previously reported falls risk assessment algorithm, and the second based on elastic net and ensemble regression models. A predictive model for falls counts in PD showed a mean R2 value of 0.43, mean error of 0.42 and a mean correlation of 30% when the results were averaged across two independent sets of PD data. The results also suggest a strong association between falls counts and a previously reported inertial sensor-based falls risk estimate. In addition, significant associations were observed between falls counts and a number of individual gait and mobility parameters. Our preliminary research suggests that the falls counts predicted from the inertial sensor data obtained during a simple walking task have the potential to be developed as a novel digital biomarker for PD, and this deserves further validation in the targeted clinical population.

International analysis on social and personal determinants of traffic violations and accidents employing logistic regression with elastic net regularization

While the number of road fatalities is declining in developed countries, it is still increasing globally, especially in middle-and low-income countries. In addition to the driver's individual awareness and attitude toward traffic safety, various factors such as the development of road infrastructure and the legal system may have a significant influence on the occurrence of traffic accidents. Thus, it is essential to consider these factors to enhance traffic safety and achieve Sustainable Development Goal (SDG) Targets of 3.6.In this study, we developed an Elastic Net Regression model to evaluate the factors that influence an individual's traffic violations and accidents based on an international questionnaire on traffic safety attitude, country fact survey data on traffic regulations, and other statistical databases. As a result, it was revealed that: i) In addition to country-level factors, the individual's attributes and attitudes toward traffic safety have an influence on the experience of traffic violations and accidents. and ii) While the same variables regarding individual attributes and attitudes are selected for both traffic violations and accidents, the selected variables relating to country factors differ between violations and accidents.

A Machine Learning Approach to Assess Differential Item Functioning in Psychometric Questionnaires Using the Elastic Net Regularized Ordinal Logistic Regression in Small Sample Size Groups.

Assessing differential item functioning (DIF) using the ordinal logistic regression (OLR) model highly depends on the asymptotic sampling distribution of the maximum likelihood (ML) estimators. The ML estimation method, which is often used to estimate the parameters of the OLR model for DIF detection, may be substantially biased with small samples. This study is aimed at proposing a new application of the elastic net regularized OLR model, as a special type of machine learning method, for assessing DIF between two groups with small samples. Accordingly, a simulation study was conducted to compare the powers and type I error rates of the regularized and nonregularized OLR models in detecting DIF under various conditions including moderate and severe magnitudes of DIF (DIF = 0.4 and 0.8), sample size (N), sample size ratio (R), scale length (I), and weighting parameter (w). The simulation results revealed that for I = 5 and regardless of R, the elastic net regularized OLR model with w = 0.1, as compared with the nonregularized OLR model, increased the power of detecting moderate uniform DIF (DIF = 0.4) approximately 35% and 21% for N = 100 and 150, respectively. Moreover, for I = 10 and severe uniform DIF (DIF = 0.8), the average power of the elastic net regularized OLR model with 0.03 ≤ w ≤ 0.06, as compared with the nonregularized OLR model, increased approximately 29.3% and 11.2% for N = 100 and 150, respectively. In these cases, the type I error rates of the regularized and nonregularized OLR models were below or close to the nominal level of 0.05. In general, this simulation study showed that the elastic net regularized OLR model outperformed the nonregularized OLR model especially in extremely small sample size groups. Furthermore, the present research provided a guideline and some recommendations for researchers who conduct DIF studies with small sample sizes.

Cardiovascular Autonomic Function Changes and Predictors During a 2-Year Physical Activity Program in Rheumatoid Arthritis: A PARA 2010 Substudy.

Background: Chronic inflammation leads to autonomic dysfunction, which may contribute to the increased risk of cardiovascular diseases (CVD) in patients with rheumatoid arthritis (RA). Exercise is known to restore autonomic nervous system (ANS) activity and particularly its parasympathetic component. A practical clinical tool to assess autonomic function, and in particular parasympathetic tone, is heart rate recovery (HRR). The aim of this substudy from the prospective PARA 2010 study was to determine changes in HRR post-maximal exercise electrocardiogram (ECG) after a 2-year physical activity program and to determine the main predictive factors associated with effects on HRR in RA.Methods: Twenty-five participants performed physiotherapist-guided aerobic and muscle-strengthening exercises for 1 year and were instructed to continue the unsupervised physical activity program autonomously in the next year. All participants were examined at baseline and at years 1 and 2 with a maximal exercise ECG on a cycle ergometer. HRR was measured at 1, 2, 3, 4, and 5 min following peak heart rate during exercise. Machine-learning algorithms with the elastic net linear regression models were performed to predict changes in HRR1 and HRR2 at 1 year and 2 years of the PARA program.Results: Mean age was 60 years, range of 41–73 years (88% women). Both HRR1 and HRR2 increased significantly from baseline to year 1 with guided physical activity and decreased significantly from year 1 to year 2 with unsupervised physical activity. Blood pressure response to exercise, low BMI, and muscular strength were the best predictors of HRR1/HRR2 increase during the first year and HRR1/HRR2 decrease during the second year of the PARA program.Conclusion: ANS activity in RA assessed by HRR was improved by guided physical activity, and machine learning allowed to identify predictors of the HRR response at the different time points. HRR could be a relevant marker of the effectiveness of physical activity recommended in patients with RA at high risk of CVD. Very inactive and/or high CVD risk RA patients may get substantial benefits from a physical activity program.

Are All Anxieties Created Equal? Stress-related Networks and Anxiety Phenotypes in Old Age

ObjectiveThe dysregulation of stress-related networks due to chronic symptoms such as severe worry and/or rumination is one of the putative pathways linking anxiety in late-life with cognitive decline and increased cardiovascular burden. Symptoms such as severe worry or rumination respond poorly to standard treatment and drive the morbidity associated with anxiety in older adults. We assessed if any of the neural networks anchored in the stress-related regions of interest (ROIs) are associated with distinct anxiety phenotypes (worry, rumination and global anxiety). MethodsWe recruited older participants (over 50 years of age) with varying levels of worry (N = 91) to undergo resting state fMRI. We computed seed-based connectivity for each ROI: the bed nucleus of the stria terminalis, the paraventricular nucleus of the hypothalamus, habenula, and amygdala. We limited our connectivity analyses to extracted regions for each seeded ROI-based network based on their canonical networks in 1,000 participants (Neurosynth). Using connectivity and clinical factors, we fit cross-validated elastic net models to predict scores on Penn State Worry Questionnaire, Rumination Subscale Questionnaire, Hamilton Anxiety Rating Scale, and Perceived Stress Scale. ResultsWe identified several distinct connectivity patterns that predict anxiety phenotypes’ severity. Greater worry was associated with greater paraventricular nucleus of the hypothalamus -subgenual anterior cingulate cortex, parahippocampal, and olfactory and amygdala-PHC connectivity. Greater global anxiety was associated with lower amygdala-superior temporal gyrus connectivity. Greater perceived stress was associated with lower amygdala-inferior temporal gyrus and amygdala-fusiform gyrus connectivity. ConclusionOur study suggests that various late-life anxiety phenotypes (worry, global anxiety, rumination) may be associated with varying functional connectivity related to stress and emotion regulation. This may aid in the development of future targeted interventions.

Sex differences in brain aging among adults with APOE4 genetic risk and family history of Alzheimer’s disease

AbstractBackgroundDetermining the effect of AD risk factors on brain aging in healthy middle‐aged and older women, compared to men, is critical for understanding whether there are sex differences in the pathways towards AD in cognitively intact, at risk adults. The goal of this study was to use an elastic net regression model to predict the effect of family history of AD [+FH] and the APOE‐e4 [+APOE4] genotype on brain age gap (BAG) in cognitively normal women and men.MethodThis cross‐sectional study used T1‐weighted structural MRIs from the Dallas Lifespan Brain Study (DLBS), South Asian Lifespan Dataset (SALD), Montreal Memory and Aging Lifespan Study (MMALS), and Pre‐symptomatic Evaluation of Experimental or Novel Treatments for Alzheimer’s Disease (PREVENT‐AD) cohorts. Data were collected from multiple sites: Dallas, USA (DLBS); Chonqing, China (SALD); Montreal, Canada (MMALS and PREVENT‐AD). 1067 cognitively normal adults (18‐89 years; 697 [65%] women) were randomly split into train and test sets. In the train set: 596 unknown FH (uFH) participants (mean [SD] age = 47.39 [18.46] years, 63% women); in the test set: 251 uFH (47.50 [19.15] years, 64% women) and 220 +FH participants (61.53 [6.04] years, 72% women). BAG, was measured as the difference between Predicted Age, estimated by the elastic net model, and Chronological Age.ResultThe age prediction model applied to the test set revealed greater BAG deviations in the +FH cohort compared to the uFH cohort (31% vs. 74% of the age variance, respectively; MAE = 8.22 years). A significant interaction was observed between FH status and Sex: +FH women had more advanced brain aging compared to +FH men. APOE genotype interacted with Sex in the +FH group: +APOE4 women had more positive BAG than +APOE4 men. In a sub‐cohort of individuals with AD risk factors, we observed sex differences in the role of modifiable factors of Body Mass Index and physical activity in preserving brain aging.ConclusionAD risk factors have more negative consequences to brain aging in cognitively normal women with family history than men with family history, but that modifiable factors can interact with genetic factors to potentially protect against brain aging.

Prenatal and postnatal exposures to four metals mixture and IQ in 6-year-old children: A prospective cohort study in South Korea.

Humans are exposed to a mixture of metals during their lifetime; however, evidence of neurotoxicity of such mixtures in critical time windows is still insufficient. We aimed to elucidate the associations of four metals mixture across multiple time points with children's intelligence quotient (IQ) in a prospective cohort study. Prenatal exposure and exposure at age 4 and 6years to four types of blood metals, namely lead, mercury, cadmium, and manganese were quantified in 502 pregnant women and their children who participated in the Environment and Development Cohort study. Children' s IQ scores were assessed using the Wechsler Intelligence Scale at age 6. Bayesian kernel machine regression (BKMR), quantile g-computation models, and elastic net (ENET) models were used to assess the associations of their blood metals mixture with IQ scores. Multivariate linear regression models indicated that postnatal blood manganese exposure at the age of 4years was significantly negatively associated with children's IQ [β=-5.99, 95% confidence interval (CI): -11.37 to-0.61]. In the multi-chemical BKMR and quantile g-computation model, statistically significant inverse associations were found between the mixture of prenatal and postnatal metals and children's IQ score (Difference in children' IQ per quartile increase: -2.83; 95% CI: -5.28, -0.38). Interestingly, we found that manganese levels at both age of 4 and 6years were contributing factors to children's IQ in the mixture models, namely, BKMR, quantile g-computation, and ENET models. Multi-pollutant mixtures of prenatal and postnatal exposures to four metals affected child IQ at 6years of age. We found a relationship between manganese exposure at both age 4, and 6years and children's IQ. Additional studies are warranted to confirm these associations and to control the exposure to different metals during pregnancy and preschool childhood.

A case-cohort study of perinatal exposure to potential endocrine disrupters and the risk of cryptorchidism in the Norwegian HUMIS study.

Exposure to endocrine-disrupting chemicals (EDCs) during the critical period of testicular descent may increase the risk of cryptorchidism and male fertility. To investigate 27 potential EDCs measured in breast milk as a proxy for perinatal exposure and the risk of cryptorchidism in a prospective cohort. The Norwegian Human Milk Study (2002-2009) enrolled 2606 mother-infant pairs, of which 1326 were mother-son pairs. In a case-cohort design, we studied 641 male infants who had 27 EDCs already quantified in milk samples: 5 organochlorine pesticides, 14 polychlorinated biphenyls (PCBs), 6 brominated flame retardants, and 2 poly- and perfluoroalkyl substances. We defined cases of congenital, recurrent, persistent and ever-reported cryptorchidism based on questionnaires mothers completed when children were 1, 6, 12 and 24months old. Variable selection via elastic net logistic regression identified the best cryptorchidism predictors while multivariable logistic regression models determined their effect estimates. The prevalence of reported congenital cryptorchidism was 6.1%, with half spontaneously descending within six months of birth, after which prevalence stabilized between 2.2 and 2.4%. The ever-reported prevalence of cryptorchidism at 1, 6, 12, or 24months was 12.2%. Elastic net models identified PCB-74 (OR=1.31, 95% CI: 1.001-1.703), PCB-114 (OR=1.36, 95% CI: 1.05-1.77), PCB-194 (OR=1.28, 95% CI: 1.03-1.53) and β-HCH (OR=1.26, 95% CI: 1.03-1.53 (per interquartile range increase in concentration of EDCs) as best predictors of congenital cryptorchidism. No EDCs were selected for either recurrent or persistent cryptorchidism, and only PCB-194 was selected by elastic net for ever-reported cryptorchidism (OR=1.23, 95% CI: 1.01-1.51), in contrast to unpenalized multivariable logistic regression, where most of the individual congeners of PCBs showed significant associations. In the largest multi-pollutant analysis to date considering potential confounding from co-exposure to other chemicals, perinatal exposure to PCB-74, PCB-114, PCB-194 and β-HCH were associated with increased odds of congenital cryptorchidism. Many PCBs may falsely be associated with cryptorchidism when assessed individually, due to confounding by highly correlated chemicals. Experimental studies are warranted to confirm our findings.

Improving Prediction of Peroxide Value of Edible Oils Using Regularized Regression Models.

We present four unique prediction techniques, combined with multiple data pre-processing methods, utilizing a wide range of both oil types and oil peroxide values (PV) as well as incorporating natural aging for peroxide creation. Samples were PV assayed using a standard starch titration method, AOCS Method Cd 8-53, and used as a verified reference method for PV determination. Near-infrared (NIR) spectra were collected from each sample in two unique optical pathlengths (OPLs), 2 and 24 mm, then fused into a third distinct set. All three sets were used in partial least squares (PLS) regression, ridge regression, LASSO regression, and elastic net regression model calculation. While no individual regression model was established as the best, global models for each regression type and pre-processing method show good agreement between all regression types when performed in their optimal scenarios. Furthermore, small spectral window size boxcar averaging shows prediction accuracy improvements for edible oil PVs. Best-performing models for each regression type are: PLS regression, 25 point boxcar window fused OPL spectral information RMSEP = 2.50; ridge regression, 5 point boxcar window, 24 mm OPL, RMSEP = 2.20; LASSO raw spectral information, 24 mm OPL, RMSEP = 1.80; and elastic net, 10 point boxcar window, 24 mm OPL, RMSEP = 1.91. The results show promising advancements in the development of a full global model for PV determination of edible oils.

Stress prediction using micro-EMA and machine learning during COVID-19 social isolation.

Accurately predicting users’ perceived stress is beneficial to aid early intervention and prevent both mental illness and physical disease during the COVID-19 pandemic. However, the existing perceived stress predicting system needs to collect a large amount of previous data for training but has a limited prediction range (i.e., next 1–2 days). Therefore, we propose a perceived stress prediction system based on the history data of micro-EMA for identifying risks 7 days earlier. Specifically, we first select and deliver an optimal set of micro-EMA questions to users every Monday, Wednesday, and Friday for reducing the burden. Then, we extract time-series features from the past micro-EMA responses and apply an Elastic net regularization model to discard redundant features. After that, selected features are fed to an ensemble prediction model for forecasting fine-grained perceived stress in the next 7 days. Experiment results show that our proposed prediction system can achieve around 4.26 (10.65% of the scale) mean absolute error for predicting the next 7 day’s PSS scores, and higher than 81% accuracy for predicting the next 7 day’s stress labels.

Abstract 11325: Lipidomics Profiling of Risk for Subclinical Atherosclerosis: A Longitudinal Study in American Indians in the Strong Heart Study

Background: Dyslipidemia is an established risk factor for atherosclerotic CVD, but full identification of circulating molecular lipids associated with atherosclerosis is still lacking in any racial/ethnic group. Objective: Prospectively identify lipid species associated with risk of atherosclerosis in American Indians. Methods: Using an untargeted LC-MS, we repeatedly measured 3,891 paired fasting plasma samples from 1,950 American Indians at two exams (~5.5 year apart) in the Strong Heart Family Study. Carotid atherosclerosis was assessed by carotid ultrasonography. Incident carotid atherosclerosis was identified if plaque was absent at baseline but detected at follow-up. Elastic net model was used to select baseline lipids or change in lipids (baseline to follow-up) associated with risk for atherosclerosis beyond traditional risk factors (TRFs) including age, sex, center, smoking, BMI, hypertension, diabetes, LDL-c and albuminuria. Multivariate analysis was used to identify lipidomic signatures associated with risk for carotid atherosclerosis. Results: Of 1,319 participants with no evident plaque at baseline (2001-2003), 270 developed incident carotid atherosclerosis at 5-year follow-up (2006-2009). Of 518 known lipids detected, altered baseline levels of 29 lipid species, largely glycerophospholipids and glycerolipids, were significantly associated with risk of atherosclerosis beyond TRFs (HR: 1.06-1.33 for risk lipids; 0.85-0.93 for protective lipids). A lipid score comprising top 5 baseline lipids improves risk prediction beyond TRFs (ROC increases from 0.79 to 0.80, P = 0.03). Longitudinal changes in 20 lipids, mainly glycerophospholipids, were significantly associated with risk of atherosclerosis after adjusting for TRFs and baseline lipids. Multivariate analysis identified distinct lipidomic signatures associated with atherosclerosis. Conclusion: Altered baseline lipids were independently associated with prevalent and incident carotid atherosclerosis in American Indians. Our findings shed light on the mechanisms linking dyslipidemia to atherosclerosis and provide instrumental data for risk stratification and therapeutic targets customized to American Indians.

Development of a Risk Prediction Model for Assessing Dental Readiness in the Canadian Armed Forces.

The establishment and sustainment of a high state of dental readiness in the Canadian Armed Forces (CAF) are the primary missions of the Royal Canadian Dental Corps. The objective of this study was to develop a risk prediction tool to estimate dental readiness in active CAF personnel. The prediction model was developed to predict the classification of non-deployable (yes/no) within 12 months (primary) and 18 months (secondary) using both dental history data (including dental attendance, restorations, root canals, and third molar status) and demographic information. Two cohorts were used for development: a recruit cohort who enrolled between April 2016 and March 2017 and a longer-serving member (LSM) cohort who had their recall dental exam between May 2014 and October 2014. Each group was followed until April 26, 2018. Elastic net logistic regression models were used to create the models. Model performance was evaluated using area under the curve, F1, and the Brier score. The recruit cohort included 2,828 individuals and the LSM cohort included 2,398 individuals. Overall, the classification of non-deployable occurred in 5.1% of the study population within 12 months and 9.6% of the population within 18 months. The models predicted the outcome with an area under the receiver operating curve of 0.77 in recruits and 0.70 in LSMs. The prediction model shows potential but its performance and usability could be further improved through the consistent collection of high quality, discretely entered, epidemiological data following standardized diagnostic terminology and coding. A recalibrated and automated version of this model could assist in decision making, resource allocation, and the enhancement of military dental readiness.

Development and validation of a prediction model for actionable aspects of frailty in the text of clinicians' encounter notes.

Frailty is a prevalent risk factor for adverse outcomes among patients with chronic lung disease. However, identifying frail patients who may benefit from interventions is challenging using standard data sources. We therefore sought to identify phrases in clinical notes in the electronic health record (EHR) that describe actionable frailty syndromes. We used an active learning strategy to select notes from the EHR and annotated each sentence for 4 actionable aspects of frailty: respiratory impairment, musculoskeletal problems, fall risk, and nutritional deficiencies. We compared the performance of regression, tree-based, and neural network models to predict the labels for each sentence. We evaluated performance with the scaled Brier score (SBS), where 1 is perfect and 0 is uninformative, and the positive predictive value (PPV). We manually annotated 155952 sentences from 326 patients. Elastic net regression had the best performance across all 4 frailty aspects (SBS 0.52, 95% confidence interval [CI] 0.49-0.54) followed by random forests (SBS 0.49, 95% CI 0.47-0.51), and multi-task neural networks (SBS 0.39, 95% CI 0.37-0.42). For the elastic net model, the PPV for identifying the presence of respiratory impairment was 54.8% (95% CI 53.3%-56.6%) at a sensitivity of 80%. Classification models using EHR notes can effectively identify actionable aspects of frailty among patients living with chronic lung disease. Regression performed better than random forest and neural network models. NLP-based models offer promising support to population health management programs that seek to identify and refer community-dwelling patients with frailty for evidence-based interventions.

BIOM-43. CROSS-PLATFORM ROBUSTNESS IN THE GLUCOCORTICOID RESPONSE PHARMACODYNAMIC BIOMARKER

Abstract The neutrophil dexamethasone methylation index (NDMI) is an algorithm-based biomarker to assess individuals’ exposures to dexamethasone, a synthetic glucocorticoid commonly administered for inflammation. Cortisol is the main endogenous glucocorticoid that controls vital processes including the immune response and lipid and carbohydrate metabolism. Variations in the NDMI score reflect individuals’ sensitivities of exposures to both exogenous and endogenous glucocorticoids, and this biomarker was trained using elastic net regression on Illumina’s most recent DNA methylation beadarray, the EPIC array, which contains 850,000 cytosine-guanine (CpG) sites. While technology for microarray research continues to advance over time, researchers are capable of conducting more comprehensive epigenome-wide association studies (EWAS). However, many studies are still run and archived using Illumina’s historical 450K platform with approximately 450,000 CpGs, and there are fewer published databases using the 850K EPIC array. To evaluate the cross-platform bioinformatic comparability, we performed elastic net regression modeling using predictors available in the 450K to train the NDMI. Among the 135 pre-surgery glioma cases from the UCSF Immune Profiles Study (IPS), NDMI scores between the 450K and 850K model were strongly correlated (r = 0.99, p &amp;lt; 0.0001). In the 311 controls from the UCSF Adult Glioma Study (AGS), similar correlations were observed (r = 0.96, p &amp;lt; 0.0001). We observe that NDMI remains a robust tool using historical 450K data and conclude that this algorithmic tool is capable of detecting the variations in individuals’ responses to dexamethasone.

Crack-Considered Elastic Net Monitoring Model of Concrete Dam Displacement

Displacement monitoring data modeling is important for evaluating the performance and health conditions of concrete dams. Conventional displacement monitoring models of concrete dams decompose the total displacement into the water pressure component, temperature component, and time-dependent component. And the crack-induced displacement is generally incorporated into the time-dependent component, thus weakening the interpretability of the model. In the practical engineering modeling, some significant explaining variables are selected while the others are eliminated by applying commonly used regression methods which occasionally show instability. This paper proposes a crack-considered elastic net monitoring model of concrete dam displacement to improve the interpretability and stability. In this model, the mathematical expression of the crack-induced displacement component is derived through the analysis of large surface crack’s effect on the concrete dam displacement to improve the interpretability of the model. Moreover, the elastic net method with better stability is used to solve the crack-considered displacement monitoring model. Sequentially, the proposed model is applied to analyze the radial displacement of a gravity arch dam. The results demonstrate that the proposed model contributes to more reasonable explaining variables’ selection and better coefficients’ estimation and also indicate better interpretability and higher predictive precision.

Machine Learning Identifies Stool pH as a Predictor of Bone Mineral Density in Healthy Multiethnic US Adults

BackgroundA variety of modifiable and nonmodifiable factors such as ethnicity, age, and diet have been shown to influence bone health. Previous studies are usually limited to analyses focused on the association of a few a priori variables or on a specific subset of the population. ObjectiveDietary, physiological, and lifestyle data were used to identify directly modifiable and nonmodifiable variables predictive of bone mineral content (BMC) and bone mineral density (BMD) in healthy US men and women using machine-learning models. MethodsRidge, lasso, elastic net, and random forest models were used to predict whole-body, femoral neck, and spine BMC and BMD in healthy US men and women ages 18–66 y, with a BMI (kg/m2) of 18–44 (n = 313), using nonmodifiable anthropometric, physiological, and demographic variables; directly modifiable lifestyle (physical activity, tobacco use) and dietary (via FFQ) variables; and variables approximating directly modifiable behavior (circulating 25-hydroxycholecalciferol and stool pH). ResultsMachine-learning models using nonmodifiable variables explained more variation in BMC and BMD (highest R2 = 0.75) compared with when using only directly modifiable variables (highest R2 = 0.11). Machine-learning models had better performance compared with multivariate linear regression, which had lower predictive value (highest R2 = 0.06) when using directly modifiable variables only. BMI, body fat percentage, height, and menstruation history were predictors of BMC and BMD. For directly modifiable features, betaine, cholesterol, hydroxyproline, menaquinone-4, dihydrophylloquinone, eggs, cheese, cured meat, refined grains, fruit juice, and alcohol consumption were predictors of BMC and BMD. Low stool pH, a proxy for fermentable fiber intake, was also predictive of higher BMC and BMD. ConclusionsModifiable factors, such as diet, explained less variation in the data compared with nonmodifiable factors, such as age, sex, and ethnicity, in healthy US men and women. Low stool pH predicted higher BMC and BMD. This trial was registered at http://www.clinicaltrials.gov as NCT02367287.

Fecal Microbiome Composition Does Not Predict Diet-Induced TMAO Production in Healthy Adults.

BackgroundTrimethylamine‐N‐oxide (TMAO) is a small molecule derived from the metabolism of dietary nutrients by gut microbes and contributes to cardiovascular disease. Plasma TMAO increases following consumption of red meat. This metabolic change is thought to be partly because of the expansion of gut microbes able to use nutrients abundant in red meat.Methods and ResultsWe used data from a randomized crossover study to estimate the degree to which TMAO can be estimated from fecal microbial composition. Healthy participants received a series of 3 diets that differed in protein source (red meat, white meat, and non‐meat), and fecal, plasma, and urine samples were collected following 4 weeks of exposure to each diet. TMAO was quantitated in plasma and urine, while shotgun metagenomic sequencing was performed on fecal DNA. While the cai gene cluster was weakly correlated with plasma TMAO (rho=0.17, P=0.0007), elastic net models of TMAO were not improved by abundances of bacterial genes known to contribute to TMAO synthesis. A global analysis of all taxonomic groups, genes, and gene families found no meaningful predictors of TMAO. We postulated that abundances of known genes related to TMAO production do not predict bacterial metabolism, and we measured choline‐ and carnitine‐trimethylamine lyase activity during fecal culture. Trimethylamine lyase genes were only weakly correlated with the activity of the enzymes they encode.ConclusionsFecal microbiome composition does not predict systemic TMAO because, in this case, gene copy number does not predict bacterial metabolic activity.RegistrationURL: https://www.clinicaltrials.gov; Unique identifier: NCT01427855.

Improved Metabolite Prediction Using Microbiome Data-Based Elastic Net Models.

Microbiome data are becoming increasingly available in large health cohorts, yet metabolomics data are still scant. While many studies generate microbiome data, they lack matched metabolomics data or have considerable missing proportions of metabolites. Since metabolomics is key to understanding microbial and general biological activities, the possibility of imputing individual metabolites or inferring metabolomics pathways from microbial taxonomy or metagenomics is intriguing. Importantly, current metabolomics profiling methods such as the HMP Unified Metabolic Analysis Network (HUMAnN) have unknown accuracy and are limited in their ability to predict individual metabolites. To address this gap, we developed a novel metabolite prediction method, and we present its application and evaluation in an oral microbiome study. The new method for predicting metabolites using microbiome data (ENVIM) is based on the elastic net model (ENM). ENVIM introduces an extra step to ENM to consider variable importance (VI) scores, and thus, achieves better prediction power. We investigate the metabolite prediction performance of ENVIM using metagenomic and metatranscriptomic data in a supragingival biofilm multi-omics dataset of 289 children ages 3–5 who were participants of a community-based study of early childhood oral health (ZOE 2.0) in North Carolina, United States. We further validate ENVIM in two additional publicly available multi-omics datasets generated from studies of gut health. We select gene family sets based on variable importance scores and modify the existing ENM strategy used in the MelonnPan prediction software to accommodate the unique features of microbiome and metabolome data. We evaluate metagenomic and metatranscriptomic predictors and compare the prediction performance of ENVIM to the standard ENM employed in MelonnPan. The newly developed ENVIM method showed superior metabolite predictive accuracy than MelonnPan when trained with metatranscriptomics data only, metagenomics data only, or both. Better metabolite prediction is achieved in the gut microbiome compared with the oral microbiome setting. We report the best-predictable compounds in all these three datasets from two different body sites. For example, the metabolites trehalose, maltose, stachyose, and ribose are all well predicted by the supragingival microbiome.