Application of quantile regression to recent genetic and -omic studies

Enhancing genomic prediction with Stacking Ensemble Learning in Arabica Coffee

Coffee Breeding programs have traditionally relied on observing plant characteristics over years, a slow and costly process. Genomic selection (GS) offers a DNA-based alternative for faster selection of superior cultivars. Stacking Ensemble Learning (SEL) combines multiple models for potentially even more accurate selection. This study explores SEL potential in coffee breeding, aiming to improve prediction accuracy for important traits [yield (YL), total number of the fruits (NF), leaf miner infestation (LM), and cercosporiosis incidence (Cer)] in Coffea Arabica. We analyzed data from 195 individuals genotyped for 21,211 single-nucleotide polymorphism (SNP) markers. To comprehensively assess model performance, we employed a cross-validation (CV) scheme. Genomic Best Linear Unbiased Prediction (GBLUP), multivariate adaptive regression splines (MARS), Quantile Random Forest (QRF), and Random Forest (RF) served as base learners. For the meta-learner within the SEL framework, various options were explored, including Ridge Regression, RF, GBLUP, and Single Average. The SEL method was able to predict the predictive ability (PA) of important traits in Coffea Arabica. SEL presented higher PA compared with those obtained for all base learner methods. The gains in PA in relation to GBLUP were 87.44% (the ratio between the PA obtained from best Stacking model and the GBLUP), 37.83%, 199.82%, and 14.59% for YL, NF, LM and Cer, respectively. Overall, SEL presents a promising approach for GS. By combining predictions from multiple models, SEL can potentially enhance the PA of GS for complex traits.

Identifying important gene signatures of BMI using network structure-aided nonparametric quantile regression

We focus on identifying genomics risk factors of higher body mass index (BMI) incorporating a priori information, such as biological pathways. However, the commonly used methods to incorporate prior information provide a model for the mean function of the outcome and rely on unmet assumptions. To address these concerns, we propose a method for nonparametric additive quantile regression with network regularization to incorporate the information encoded by known networks. To account for nonlinear associations, we approximate the unknown additive functional effect of each predictor with the expansion of a B-spline basis. We implement the group Lasso penalty to obtain a sparse model. We define the network-constrained penalty by the totalℓ 2 $$ {\ell}_2 $$ norm of the difference between the effect functions of any two linked genes in the known network. We further propose an efficient computation procedure to solve the optimization problem that arises in our model. Simulation studies show that our proposed method performs well in identifying more truly associated genes and less falsely associated genes than alternative approaches. We apply the proposed method to analyze the microarray gene-expression dataset in the Framingham Heart Study and identify several 75 percentile BMI associated genes. In conclusion, our proposed approach efficiently identifies the outcome-associated variables in a nonparametric additive quantile regression framework by leveraging known network information.

TENSOR QUANTILE REGRESSION WITH APPLICATION TO ASSOCIATION BETWEEN NEUROIMAGES AND HUMAN INTELLIGENCE

Human intelligence is usually measured by well-established psychometric tests through a series of problem solving. The recorded cognitive scores are continuous but usually heavy-tailed with potential outliers and violating the normality assumption. Meanwhile, magnetic resonance imaging (MRI) provides an unparalleled opportunity to study brain structures and cognitive ability. Motivated by association studies between MRI images and human intelligence, we propose a tensor quantile regression model, which is a general and robust alternative to the commonly used scalar-on-image linear regression. Moreover, we take into account rich spatial information of brain structures, incorporating low-rankness and piece-wise smoothness of imaging coefficients into a regularized regression framework. We formulate the optimization problem as a sequence of penalized quantile regressions with a generalized Lasso penalty based on tensor decomposition, and develop a computationally efficient alternating direction method of multipliers algorithm (ADMM) to estimate the model components. Extensive numerical studies are conducted to examine the empirical performance of the proposed method and its competitors. Finally, we apply the proposed method to a large-scale important dataset: the Human Connectome Project. We find that the tensor quantile regression can serve as a prognostic tool to assess future risk of cognitive impairment progression. More importantly, with the proposed method, we are able to identify the most activated brain subregions associated with quantiles of human intelligence. The prefrontal and anterior cingulate cortex are found to be mostly associated with lower and upper quantile of fluid intelligence. The insular cortex associated with median of fluid intelligence is a rarely reported region.

Age-related DNA methylation analysis for forensic age estimation using post-mortem blood samples from Japanese individuals

As one of external visible characteristics (EVCs) in forensic phenotyping, age estimation is essential to providing additional information about a sample donor. With the development of epigenetics, age-related DNA methylation may be used as a reliable predictor of age estimation. With the aim of building a feasible age estimation model for Japanese individuals, 53 CpG sites distributed between 11 candidate genes were selected from previous studies. The DNA methylation level of each target CpG site was identified and measured on a massive parallel platform (synthesis by sequencing, Illumina, California, United States) from 60 forensic blood samples during the initial training phase. Multiple linear regression and quantile regression analyses were later performed to build linear and quantile age estimation models, respectively. Four CpG sites on four genes- ASPA, ELOVL2, ITGA2B, and PDE4C -, were found to be highly correlated with chronological age in DNA samples from Japanese individuals (|R| > 0.75). Subsequently, an independent validation dataset (n = 30) was used to verify and evaluate the performance of the two models. Comparison of mean absolute deviation (MAD) with other indicators showed that both models provide accurate age predictions (MAD: linear = 6.493 years; quantile = 6.243 years). The quantile model, however, can provide the changeable prediction intervals that grow wider with increasing age, and this tendency is consistent with the natural aging process in humans. Hence, the quantile model is recommended in this study.

Genetic Susceptibility Determines β-Cell Function and Fasting Glycemia Trajectories Throughout Childhood: A 12-Year Cohort Study (EarlyBird 76)

OBJECTIVE

Previous studies suggested that childhood prediabetes may develop prior to obesity and be associated with relative insulin deficiency. We proposed that the insulin-deficient phenotype is genetically determined and tested this hypothesis by longitudinal modeling of insulin and glucose traits with diabetes risk genotypes in the EarlyBird cohort.

RESEARCH DESIGN AND METHODS

EarlyBird is a nonintervention prospective cohort study that recruited 307 healthy U.K. children at 5 years of age and followed them throughout childhood. We genotyped 121 single nucleotide polymorphisms (SNPs) previously associated with diabetes risk, identified in the adult population. Association of SNPs with fasting insulin and glucose and HOMA indices of insulin resistance and β-cell function, available from 5 to 16 years of age, were tested. Association analysis with hormones was performed on selected SNPs.

RESULTS

Several candidate loci influenced the course of glycemic and insulin traits, including rs780094 (GCKR), rs4457053 (ZBED3), rs11257655 (CDC123), rs12779790 (CDC123 and CAMK1D), rs1111875 (HHEX), rs7178572 (HMG20A), rs9787485 (NRG3), and rs1535500 (KCNK16). Some of these SNPs interacted with age, the growth hormone-IGF-1 axis, and adrenal and sex steroid activity.

CONCLUSIONS

The findings that genetic markers influence both elevated and average courses of glycemic traits and β-cell function in children during puberty independently of BMI are a significant step toward early identification of children at risk for diabetes. These findings build on our previous observations that pancreatic β-cell defects predate insulin resistance in the onset of prediabetes. Understanding the mechanisms of interactions among genetic factors, puberty, and weight gain would allow the development of new and earlier disease-management strategies in children.

Contribution of Four Polymorphisms in Renin-Angiotensin-Aldosterone-Related Genes to Hypertension in a Thai Population

Introduction

The roles of genes in the renin-angiotensin-aldosterone system (RAAS) in hypertension, including angiotensin-converting enzyme (ACE), angiotensinogen (AGT), angiotensin II receptor type 1 (AGTR1), and aldosterone synthase (CYP11B2), have been widely studied across different ethnicities, but there has been no such investigation in Thai population.

Materials and Methods

Using 4,150 Thais recorded in the Electricity Generating Authority of Thailand (EGAT) study, we examined the association of rs1799752, rs699, rs5186, and rs1799998 located in or near ACE, AGT, AGTR1, and CYP11B2 genes in hypertension. We investigated their roles in hypertension using multivariate logistic regression and further examined their roles in blood pressure (BP) using quantile regression. Sex, age, and BMI were adjusted as potential confounders.

Results

We did not observe associations between hypertension and rs1799752 (P=0.422), rs699 (P=0.36), rs5186 (P=0.49), and rs1799998 (P=0.71). No evidence of association between these SNPs and BP was found across an entire distribution. A nonlinear relationship between age and BP was observed.

Conclusion

In Thai population, our study showed no evidence of association between RAAS-related genes and hypertension. While our study is the first and largest study to investigate the role of RAAS-related genes in hypertension in Thai population, restricted statistical power due to limited sample size is a limitation.

Is a mixture of arbuscular mycorrhizal fungi better for plant growth than single-species inoculants?

Inoculation of arbuscular mycorrhizal fungi (AMF) as plant growth promoters has mostly been conducted using single-species inoculum. In this study, we investigated whether co-inoculation of different native AMF species induced an improvement of plant growth in an ultramafic soil. We analyzed the effects of six species of AMF from a New Caledonian ultramafic soil on plant growth and nutrition, using mono-inoculations and mixtures comprising different numbers of AMF species, in a greenhouse experiment. The endemic Metrosideros laurifolia was used as a host plant. Our results suggest that, when the plant faced multiple abiotic stress factors (nutrient deficiencies and high concentrations of different heavy metals), co-inoculation of AMF belonging to different families was more efficient than mono-inoculation in improving biomass, mineral nutrition, Ca/Mg ratio, and tolerance to heavy metals of plants in ultramafic soil. This performance suggested functional complementarity between distantly related AMF. Our findings will have important implications for restoration ecology and mycorrhizal biotechnology applied to ultramafic soils.

Using quantile regression for fitting lactation curve in dairy cows

The main objective of this study was to compare the performance of different 'nonlinear quantile regression' models evaluated at the τth quantile (0·25, 0·50, and 0·75) of milk production traits and somatic cell score (SCS) in Iranian Holstein dairy cows. Data were collected by the Animal Breeding Center of Iran from 1991 to 2011, comprising 101 051 monthly milk production traits and SCS records of 13 977 cows in 183 herds. Incomplete gamma (Wood), exponential (Wilmink), Dijkstra and polynomial (Ali & Schaeffer) functions were implemented in the quantile regression. Residual mean square, Akaike information criterion and log-likelihood from different models and quantiles indicated that in the same quantile, the best models were Wilmink for milk yield, Dijkstra for fat percentage and Ali & Schaeffer for protein percentage. Over all models the best model fit occurred at quantile 0·50 for milk yield, fat and protein percentage, whereas, for SCS the 0·25th quantile was best. The best model to describe SCS was Dijkstra at quantiles 0·25 and 0·50, and Ali & Schaeffer at quantile 0·75. Wood function had the worst performance amongst all traits. Quantile regression is specifically appropriate for SCS which has a mixed multimodal distribution.

Treating a broader range of depressed adolescents with combined therapy

BACKGROUND

Traditional statistical analyses of clinical trials encompass the central tendency of outcomes and, hence, are restricted to a treatment's average effectiveness. Our aim was to get a more complete picture of the effectiveness of standard treatment options for adolescent depression, by analyzing treatment effects across low, middle, and high levels of response.

METHODS

Secondary data analysis was performed of the Treatment for Adolescents with Depression Study (TADS, ClinicalTrials.gov, NCT00006286), a randomized controlled trial comparing fluoxetine (FLX), cognitive-behavioral therapy (CBT), and their combination (COMB) against placebo treating adolescents with major depression (n = 439). The proportional change from baseline to week 12 in the Children's Depression Rating Scale-Revised was used as an index of response. Response levels were analyzed via quantile regression models, thereby estimating treatment effects across the entire response level distribution, adjusted for baseline depression, study site, and patients' treatment expectancies.

RESULTS

Whereas CBT was no more effective than placebo across response levels, COMB was more effective than FLX in that its quantile treatment effects were both larger in magnitude and spread out across a broader range of response levels, including the low end of the response level distribution. Cohen's d of the difference was 1.39 (95% confidence interval 1.33-1.45).

LIMITATIONS

Ad-hoc analysis using data from a trial that was not originally designed to accommodate such analysis.

CONCLUSION

The combination of cognitive-behavioral therapy and fluoxetine was more effective than either treatment used alone, not just in average effectiveness, but in the breadth of patients in whom it was effective.

Maltreatment, the Oxytocin Receptor Gene, and Conduct Problems Among Male and Female Teenagers

The oxytocin receptor gene (OXTR) influences human behavior. The G allele of OXTR rs53576 has been associated with both prosocial and maladaptive behaviors but few studies have taken account of environmental factors. The present study determined whether the association of childhood maltreatment with conduct problems was modified by OXTR rs53576 genotypes. In a general population sample of 1591 teenagers, conduct problems as well as maltreatment were measured by self-report. DNA was extracted from saliva samples. In males, there was a significant positive association between maltreatment and conduct problems independent of the genotype. In females, among G allele carriers, the level of conduct problems was significantly higher among those who had been maltreated as compared to those not maltreated. By contrast, among female AA carriers, conduct problems did not vary between those who were, and who were not, maltreated. The results indicate that OXTR rs53576 plays a role in antisocial behavior in females such that the G allele confers vulnerability for antisocial behavior if they experience maltreatment, whereas the A allele has a protective effect.

Quantile regression for genome-wide association study of flowering time-related traits in common bean

Flowering is an important agronomic trait. Quantile regression (QR) can be used to fit models for all portions of a probability distribution. In Genome-wide association studies (GWAS), QR can estimate SNP (Single Nucleotide Polymorphism) effects on each quantile of interest. The objectives of this study were to estimate genetic parameters and to use QR to identify genomic regions for phenological traits (Days to first flower-DFF; Days for flowering-DTF; Days to end of flowering-DEF) in common bean. A total of 80 genotypes of common beans, with 3 replicates were raised at 4 locations and seasons. Plants were genotyped for 384 SNPs. Traditional single-SNP and 9 QR models, ranging from equally spaced quantiles (τ) 0.1 to 0.9, were used to associate SNPs to phenotype. Heritabilities were moderate high, ranging from 0.32 to 0.58. Genetic and phenotypic correlations were all high, averaging 0.66 and 0.98, respectively. Traditional single-SNP GWAS model was not able to find any SNP-trait association. On the other hand, when using QR methodology considering one extreme quantile (τ = 0.1) we found, respectively 1 and 7, significant SNPs associated for DFF and DTF. Significant SNPs were found on Pv01, Pv02, Pv03, Pv07, Pv10 and Pv11 chromosomes. We investigated potential candidate genes in the region around these significant SNPs. Three genes involved in the flowering pathways were identified, including Phvul.001G214500, Phvul.007G229300 and Phvul.010G142900.1 on Pv01, Pv07 and Pv10, respectively. These results indicate that GWAS-based QR was able to enhance the understanding on genetic architecture of phenological traits (DFF and DTF) in common bean.

The trophoblast cell surface antigen 2 and miR-125b axis in urothelial bladder cancer

Human trophoblast cell surface antigen 2 (Trop-2) is a 40-kDa transmembrane glycoprotein that was first identified as a marker of human trophoblast cells. Trop-2 acts on cell proliferation, adhesion, and migration by activating a number of intracellular signalling pathways. Elevated Trop-2 expression has been demonstrated in several types of cancer and correlated with aggressiveness and poor prognosis. Since no data are available on Trop-2 in bladder cancer (BC), the purpose of the study was to determine its levels in tissue specimens from normal individuals and patients with BC at different stages. Moreover, since according to recent evidence Trop-2 is a miR-125b target, miR-125b expression was also assessed in tissue specimens. Finally, the effect of the Trop-2/miR-125b axis on the proliferation and migration of BC cells was evaluated in vitro.

The Trop-2/miR-125b axis was seen to be differentially expressed in normal urothelium, non-invasive BC and invasive BC tissue. Significant miR-125b down-regulation was associated with a significant increase in Trop-2 protein levels in BC tissue and correlated with disease severity. In vitro analysis confirmed the role of miR-125b in down-modulation of Trop-2 protein levels and showed that Trop-2/miR-125b axis affects cellular proliferation in bladder tissue.

In conclusion, our findings highlight a role for the Trop-2/miR-125b axis in BC progression and suggest Trop-2 and miR-125b as diagnostic/prognostic marker candidates as well as druggable targets for innovative therapeutic approaches.

Acute toxicities of pharmaceuticals toward green algae. mode of action, biopharmaceutical drug disposition classification system and quantile regression models

The acute toxicities of 36 pharmaceuticals towards green algae were estimated from a set of quantile regression models representing the first global quantitative structure-activity relationships. The selection of these pharmaceuticals was based on their predicted environmental concentrations. An agreement between the estimated values and the observed acute toxicity values was found for several families of pharmaceuticals, in particular, for antidepressants. A recent classification (BDDCS) of drugs based on ADME properties (Absorption, Distribution, Metabolism and Excretion) was clearly correlated with the acute ecotoxicities towards algae. Over-estimation of toxicity from our QSAR models was observed for classes 2, 3 and 4 whereas our model results were in agreement for the class 1 pharmaceuticals. Clarithromycin, a class 3 antibiotic characterized by weak metabolism and high solubility, was the most toxic to algae (molecular stability and presence in surface water).

Genetics of Plasminogen Activator Inhibitor-1 (PAI-1) in a Ghanaian Population

Plasminogen activator inhibitor 1 (PAI-1), a major modulator of the fibrinolytic system, is an important factor in cardiovascular disease (CVD) susceptibility and severity. PAI-1 is highly heritable, but the few genes associated with it explain only a small portion of its variation. Studies of PAI-1 typically employ linear regression to estimate the effects of genetic variants on PAI-1 levels, but PAI-1 is not normally distributed, even after transformation. Therefore, alternative statistical methods may provide greater power to identify important genetic variants. Additionally, most genetic studies of PAI-1 have been performed on populations of European descent, limiting the generalizability of their results. We analyzed >30,000 variants for association with PAI-1 in a Ghanaian population, using median regression, a non-parametric alternative to linear regression. Three variants associated with median PAI-1, the most significant of which was in the gene arylsulfatase B (ARSB) (p = 1.09 x 10⁻⁷). We also analyzed the upper quartile of PAI-1, the most clinically relevant part of the distribution, and found 19 SNPs significantly associated in this quartile. Of note an association was found in period circadian clock 3 (PER3). Our results reveal novel associations with median and elevated PAI-1 in an understudied population. The lack of overlap between the two analyses indicates that the genetic effects on PAI-1 are not uniform across its distribution. They also provide evidence of the generalizability of the circadian pathway’s effect on PAI-1, as a recent meta-analysis performed in Caucasian populations identified another circadian clock gene (ARNTL).

Quantile regression model for a diverse set of chemicals: application to acute toxicity for green algae

The potential of quantile regression (QR) and quantile support vector machine regression (QSVMR) was analyzed for the definitions of quantitative structure-activity relationship (QSAR) models associated with a diverse set of chemicals toward a particular endpoint. This study focused on a specific sensitive endpoint (acute toxicity to algae) for which even a narcosis QSAR model is not actually clear. An initial dataset including more than 401 ecotoxicological data for one species of algae (Selenastrum capricornutum) was defined. This set corresponds to a large sample of chemicals ranging from classical organic chemicals to pesticides. From this original data set, the selection of the different subsets was made in terms of the notion of toxic ratio (TR), a parameter based on the ratio between predicted and experimental values. The robustness of QR and QSVMR to outliers was clearly observed, thus demonstrating that this approach represents a major interest for QSAR associated with a diverse set of chemicals. We focused particularly on descriptors related to molecular surface properties.