A strategy for high-dimensional multivariable analysis classifies childhood asthma phenotypes from genetic, immunological, and environmental factors.

BACKGROUND: Associations between childhood asthma phenotypes and genetic, immunological, and environmental factors have been previously established. Yet, strategies to integrate high-dimensional risk factors from multiple distinct data sets, and thereby increase the statistical power of analyses, have been hampered by a preponderance of missing data and lack of methods to accommodate them. METHODS: We assembled questionnaire, diagnostic, genotype, microarray, RT-qPCR, flow cytometry, and cytokine data (referred to as data modalities) to use as input factors for a classifier that could distinguish healthy children, mild-to-moderate allergic asthmatics, and nonallergic asthmatics. Based on data from 260 German children aged 4-14 from our university outpatient clinic, we built a novel multilevel prediction approach for asthma outcome which could deal with a present complex missing data structure. RESULTS: The optimal learning method was boosting based on all data sets, achieving an area underneath the receiver operating characteristic curve (AUC) for three classes of phenotypes of 0.81 (95%-confidence interval (CI): 0.65-0.94) using leave-one-out cross-validation. Besides improving the AUC, our integrative multilevel learning approach led to tighter CIs than using smaller complete predictor data sets (AUC = 0.82 [0.66-0.94] for boosting). The most important variables for classifying childhood asthma phenotypes comprised novel identified genes, namely PKN2 (protein kinase N2), PTK2 (protein tyrosine kinase 2), and ALPP (alkaline phosphatase, placental). CONCLUSION: Our combination of several data modalities using a novel strategy improved classification of childhood asthma phenotypes but requires validation in external populations. The generic approach is applicable to other multilevel data-based risk prediction settings, which typically suffer from incomplete data.

Genetic scores to stratify risk of developing multiple islet autoantibodies and type 1 diabetes: A prospective study in children.

BACKGROUND: Around 0.3% of newborns will develop autoimmunity to pancreatic beta cells in childhood and subsequently develop type 1 diabetes before adulthood. Primary prevention of type 1 diabetes will require early intervention in genetically at-risk infants. The objective of this study was to determine to what extent genetic scores (two previous genetic scores and a merged genetic score) can improve the prediction of type 1 diabetes. METHODS AND FINDINGS: The Environmental Determinants of Diabetes in the Young (TEDDY) study followed genetically at-risk children at 3- to 6-monthly intervals from birth for the development of islet autoantibodies and type 1 diabetes. Infants were enrolled between 1 September 2004 and 28 February 2010 and monitored until 31 May 2016. The risk (positive predictive value) for developing multiple islet autoantibodies (pre-symptomatic type 1 diabetes) and type 1 diabetes was determined in 4,543 children who had no first-degree relatives with type 1 diabetes and either a heterozygous HLA DR3 and DR4-DQ8 risk genotype or a homozygous DR4-DQ8 genotype, and in 3,498 of these children in whom genetic scores were calculated from 41 single nucleotide polymorphisms. In the children with the HLA risk genotypes, risk for developing multiple islet autoantibodies was 5.8% (95% CI 5.0%-6.6%) by age 6 years, and risk for diabetes by age 10 years was 3.7% (95% CI 3.0%-4.4%). Risk for developing multiple islet autoantibodies was 11.0% (95% CI 8.7%-13.3%) in children with a merged genetic score of >14.4 (upper quartile; n = 907) compared to 4.1% (95% CI 3.3%-4.9%, P < 0.001) in children with a genetic score of ≤14.4 (n = 2,591). Risk for developing diabetes by age 10 years was 7.6% (95% CI 5.3%-9.9%) in children with a merged score of >14.4 compared with 2.7% (95% CI 1.9%-3.6%) in children with a score of ≤14.4 (P < 0.001). Of 173 children with multiple islet autoantibodies by age 6 years and 107 children with diabetes by age 10 years, 82 (sensitivity, 47.4%; 95% CI 40.1%-54.8%) and 52 (sensitivity, 48.6%, 95% CI 39.3%-60.0%), respectively, had a score >14.4. Scores were higher in European versus US children (P = 0.003). In children with a merged score of >14.4, risk for multiple islet autoantibodies was similar and consistently >10% in Europe and in the US; risk was greater in males than in females (P = 0.01). Limitations of the study include that the genetic scores were originally developed from case-control studies of clinical diabetes in individuals of mainly European decent. It is, therefore, possible that it may not be suitable to all populations. CONCLUSIONS: A type 1 diabetes genetic score identified infants without family history of type 1 diabetes who had a greater than 10% risk for pre-symptomatic type 1 diabetes, and a nearly 2-fold higher risk than children identified by high-risk HLA genotypes alone. This finding extends the possibilities for enrolling children into type 1 diabetes primary prevention trials.

Allele-specific methylation of type 1 diabetes susceptibility genes.

The susceptibility to autoimmune diseases is influenced by genes encoding major histocompatibility complex (MHC) proteins. By examining the epigenetic methylation maps of cord blood samples, we found marked differences in the methylation status of CpG sites within the MHC genes (cis-metQTLs) between carriers of the type 1 diabetes risk haplotypes HLA-DRB1*03-DQA1*0501-DQB1*0201 (DR3-DQ2) and HLA-DRB1*04-DQA1*0301-DQB1*0302 (DR4-DQ8). These differences were found in children and adults, and were accompanied by reduced HLA-DR protein expression in immune cells with the HLA-DR3-DQ2 haplotype. Extensive cis-metQTLs were identified in all 45 immune and non-immune type 1 diabetes susceptibility genes analyzed in this study. We observed and validated a novel association between the methylation status of CpG sites within the LDHC gene and the development of insulin autoantibodies in early childhood in children who are carriers of the highest type 1 diabetes risk genotype. Functionally relevant epigenetic changes in susceptibility genes may represent therapeutic targets for type 1 diabetes.

Characterization of missing values in untargeted MS-based metabolomics data and evaluation of missing data handling strategies.

BACKGROUND: Untargeted mass spectrometry (MS)-based metabolomics data often contain missing values that reduce statistical power and can introduce bias in biomedical studies. However, a systematic assessment of the various sources of missing values and strategies to handle these data has received little attention. Missing data can occur systematically, e.g. from run day-dependent effects due to limits of detection (LOD); or it can be random as, for instance, a consequence of sample preparation. METHODS: We investigated patterns of missing data in an MS-based metabolomics experiment of serum samples from the German KORA F4 cohort (n = 1750). We then evaluated 31 imputation methods in a simulation framework and biologically validated the results by applying all imputation approaches to real metabolomics data. We examined the ability of each method to reconstruct biochemical pathways from data-driven correlation networks, and the ability of the method to increase statistical power while preserving the strength of established metabolic quantitative trait loci. RESULTS: Run day-dependent LOD-based missing data accounts for most missing values in the metabolomics dataset. Although multiple imputation by chained equations performed well in many scenarios, it is computationally and statistically challenging. K-nearest neighbors (KNN) imputation on observations with variable pre-selection showed robust performance across all evaluation schemes and is computationally more tractable. CONCLUSION: Missing data in untargeted MS-based metabolomics data occur for various reasons. Based on our results, we recommend that KNN-based imputation is performed on observations with variable pre-selection since it showed robust results in all evaluation schemes.

Prediction of type 1 diabetes using a genetic risk model in the Diabetes Autoimmunity Study in the Young.

BACKGROUND: Genetic predisposition for type 1 diabetes (T1D) is largely determined by human leukocyte antigen (HLA) genes; however, over 50 other genetic regions confer susceptibility. We evaluated a previously reported 10-factor weighted model derived from the Type 1 Diabetes Genetics Consortium to predict the development of diabetes in the Diabetes Autoimmunity Study in the Young (DAISY) prospective cohort. Performance of the model, derived from individuals with first-degree relatives (FDR) with T1D, was evaluated in DAISY general population (GP) participants as well as FDR subjects. METHODS: The 10-factor weighted risk model (HLA, PTPN22 , INS , IL2RA , ERBB3 , ORMDL3 , BACH2 , IL27 , GLIS3 , RNLS ), 3-factor model (HLA, PTPN22, INS ), and HLA alone were compared for the prediction of diabetes in children with complete SNP data (n = 1941). RESULTS: Stratification by risk score significantly predicted progression to diabetes by Kaplan-Meier analysis (GP: P = .00006; FDR: P = .0022). The 10-factor model performed better in discriminating diabetes outcome than HLA alone (GP, P = .03; FDR, P = .01). In GP, the restricted 3-factor model was superior to HLA (P = .03), but not different from the 10-factor model (P = .22). In contrast, for FDR the 3-factor model did not show improvement over HLA (P = .12) and performed worse than the 10-factor model (P = .02) CONCLUSIONS: We have shown a 10-factor risk model predicts development of diabetes in both GP and FDR children. While this model was superior to a minimal model in FDR, it did not confer improvement in GP. Differences in model performance in FDR vs GP children may lead to important insights into screening strategies specific to these groups.

Peptide serum markers in islet autoantibody-positive children.

AIMS/HYPOTHESIS: We sought to identify minimal sets of serum peptide signatures as markers for islet autoimmunity and predictors of progression rates to clinical type 1 diabetes in a case-control study. METHODS: A double cross-validation approach was applied to first prioritise peptides from a shotgun proteomic approach in 45 islet autoantibody-positive and -negative children from the BABYDIAB/BABYDIET birth cohorts. Targeted proteomics for 82 discriminating peptides were then applied to samples from another 140 children from these cohorts. RESULTS: A total of 41 peptides (26 proteins) enriched for the functional category lipid metabolism were significantly different between islet autoantibody-positive and autoantibody-negative children. Two peptides (from apolipoprotein M and apolipoprotein C-IV) were sufficient to discriminate autoantibody-positive from autoantibody-negative children. Hepatocyte growth factor activator, complement factor H, ceruloplasmin and age predicted progression time to type 1 diabetes with a significant improvement compared with age alone. CONCLUSION/INTERPRETATION: Distinct peptide signatures indicate islet autoimmunity prior to the clinical manifestation of type 1 diabetes and enable refined staging of the presymptomatic disease period.

Multicentre Performance Evaluation of the Elecsys Anti-SARS-CoV-2 Immunoassay as an Aid in Determining Previous Exposure to SARS-CoV-2.

INTRODUCTION: We performed a multicentre evaluation of the Elecsys® Anti-SARS-CoV-2 immunoassay (Roche Diagnostics), an assay utilising a recombinant protein representing the nucleocapsid (N) antigen, for the in vitro qualitative detection of antibodies to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). METHODS: Specificity was evaluated using serum/plasma samples from blood donors and routine diagnostic specimens collected before September 2019 (i.e., presumed negative for SARS-CoV-2-specific antibodies); sensitivity was evaluated using samples from patients with polymerase chain reaction (PCR)-confirmed SARS-CoV-2 infection. Point estimates and 95% confidence intervals (CIs) were calculated. Method comparison was performed versus commercially available assays. RESULTS: Overall specificity for the Elecsys Anti-SARS-CoV-2 immunoassay (n = 9575) was 99.85% (95% CI 99.75-99.92): blood donors (n = 6714; 99.82%), routine diagnostic specimens (n = 2861; 99.93%), pregnant women (n = 2256; 99.91%), paediatric samples (n = 205; 100.00%). The Elecsys Anti-SARS-CoV-2 immunoassay demonstrated significantly higher specificity versus LIAISON SARS-CoV-2 S1/S2 IgG (99.71% vs. 98.48%), EUROIMMUN Anti-SARS-CoV-2 IgG (100.00% vs. 94.87%), ADVIA Centaur SARS-CoV-2 Total (100.00% vs. 87.32%) and iFlash SARS-CoV-2 IgM (100.00% vs. 99.58%) assays, and comparable specificity to ARCHITECT SARS-CoV-2 IgG (99.75% vs. 99.65%) and iFlash SARS-CoV-2 IgG (100.00% vs. 100.00%) assays. Overall sensitivity for Elecsys Anti-SARS-CoV-2 immunoassay samples drawn at least 14 days post-PCR confirmation (n = 219) was 93.61% (95% CI 89.51-96.46). No statistically significant differences in sensitivity were observed between the Elecsys Anti-SARS-CoV-2 immunoassay versus EUROIMMUN Anti-SARS-CoV-2 IgG (90.32% vs. 95.16%) and ARCHITECT SARS-CoV-2 IgG (84.81% vs. 87.34%) assays. The Elecsys Anti-SARS-CoV-2 immunoassay showed significantly lower sensitivity versus ADVIA Centaur SARS-CoV-2 Total (85.19% vs. 95.06%) and iFlash SARS-CoV-2 IgG (86.25% vs. 93.75%) assays, but significantly higher sensitivity versus the iFlash SARS-CoV-2 IgM assay (86.25% vs. 33.75%). CONCLUSION: The Elecsys Anti-SARS-CoV-2 immunoassay demonstrated very high specificity and high sensitivity in samples collected at least 14 days post-PCR confirmation of SARS-CoV-2 infection, supporting its use to aid in determination of previous exposure to SARS-CoV-2.

The 99th percentile upper reference limit for the 5th generation cardiac troponin T assay in the United States.

BACKGROUND: Determining diagnostic thresholds for cardiac troponin assays is key to interpreting their clinical performance. We describe the calculation of 99th percentile upper reference limits (URLs) for the Elecsys® Troponin T Gen 5 (TnT Gen 5) assay. METHODS: Plasma and serum samples from healthy US participants were prospectively evaluated using TnT Gen 5 Short Turn Around Time and 18-min assays on cobas e 411 and cobas e 601 analyzers (Roche Diagnostics); with, up to 8 TnT Gen 5 results per participant. RESULTS: A total of 10,402 TnT Gen 5 results from 1301 participants were included (50.4% female). Across 9 calculation methods, overall 99th percentile URL was 19.2 ng/l (females, 13.5-13.6 ng/l; males, 21.4-22.2 ng/l). Across different sample/assay/analyzer combinations, overall 99th percentile URLs ranged from 18.4-20.2 ng/l. Median TnT Gen 5 results increased with age, were higher in males, and ranged from 3.0-3.7 ng/l across races/ethnicities and from 3.0-3.6 ng/l across body mass index (BMI) classes. Applying additional exclusion criteria (N-terminal pro-brain natriuretic peptide, BMI and estimated glomerular filtration rate) resulted in lower 99th percentile URLs (overall, 16.9 ng/l; females, 11.8 ng/l; males, 18.5 ng/l). CONCLUSION: Our findings facilitate the interpretation of TnT Gen 5 results in US clinical practice.

Analytical performance evaluation of the Elecsys® Troponin T Gen 5 STAT assay.

BACKGROUND: We report the analytical performance of the Elecsys® Troponin T Gen 5 STAT (TnT Gen 5 STAT; Roche Diagnostics) assay. METHODS: Measuring limits/ranges were determined in lithium-heparin plasma samples per Clinical and Laboratory Standards Institute (CLSI) EP17-A2. Precision was evaluated per CLSI EP05-A2 using lithium-heparin plasma/quality control samples on cobas e 411/cobas e 601 analyzers; two duplicated runs per day for 21 days (n = 84). Cross-reactivity with other troponin forms and interference from endogenous substances/drugs was tested; recovery criterion for no cross-reactivity was within ±10%. RESULTS: Coefficients of variation (CV) for repeatability/intermediate precision were 0.7-5.6%/1.4-10.3% (cobas e 411; mean cardiac troponin T [cTnT]: 7.3-9341 ng/L) and 0.7-3.0%/1.5-6.4% (cobas e 601; mean cTnT: 7.4-9455 ng/L). There was no cross-reactivity with skeletal muscle troponin T (≤ 10,000 ng/L), skeletal muscle troponin I (≤ 100,000 ng/L), cardiac troponin I (≤ 10,000 ng/L), or human troponin C (≤ 80,000 ng/L). No interference was observed with biotin (≤ 20 ng/mL) or 34 drugs. CONCLUSION: The TnT Gen 5 STAT assay demonstrated a CV of <10% at the 99th percentile upper reference limit, meeting precision requirements (Fourth Universal Definition of Myocardial Infarction) for high-sensitivity troponin assays.

Unbiased Prediction and Feature Selection in High-Dimensional Survival Regression.

With widespread availability of omics profiling techniques, the analysis and interpretation of high-dimensional omics data, for example, for biomarkers, is becoming an increasingly important part of clinical medicine because such datasets constitute a promising resource for predicting survival outcomes. However, early experience has shown that biomarkers often generalize poorly. Thus, it is crucial that models are not overfitted and give accurate results with new data. In addition, reliable detection of multivariate biomarkers with high predictive power (feature selection) is of particular interest in clinical settings. We present an approach that addresses both aspects in high-dimensional survival models. Within a nested cross-validation (CV), we fit a survival model, evaluate a dataset in an unbiased fashion, and select features with the best predictive power by applying a weighted combination of CV runs. We evaluate our approach using simulated toy data, as well as three breast cancer datasets, to predict the survival of breast cancer patients after treatment. In all datasets, we achieve more reliable estimation of predictive power for unseen cases and better predictive performance compared to the standard CoxLasso model. Taken together, we present a comprehensive and flexible framework for survival models, including performance estimation, final feature selection, and final model construction. The proposed algorithm is implemented in an open source R package (SurvRank) available on CRAN.