Bias correction for estimated QTL effects using the penalized maximum likelihood method

Zhang, J; Yue, C. Patrick; Ym, Zhang

doi:10.1038/hdy.2011.86

Cited by 10 publications

(9 citation statements)

References 35 publications

(54 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are several methods available in the estimation of parameters in model (2), e.g., penalized maximum likelihood [ 9 , 82 ], empirical Bayesian [ 10 ], hierarchical generalized linear model [ 83 , 84 ]. Here we adopt empirical Bayesian, for technical detail the reader is referred to the original study of Xu [ 10 ].…”

Section: Methodsmentioning

confidence: 99%

Interacted QTL Mapping in Partial NCII Design Provides Evidences for Breeding by Design

Zhao

et al. 2015

PLoS ONE

View full text Add to dashboard Cite

The utilization of heterosis in rice, maize and rapeseed has revolutionized crop production. Although elite hybrid cultivars are mainly derived from the F1 crosses between two groups of parents, named NCII mating design, little has been known about the methodology of how interacted effects influence quantitative trait performance in the population. To bridge genetic analysis with hybrid breeding, here we integrated an interacted QTL mapping approach with breeding by design in partial NCII mating design. All the potential main and interacted effects were included in one full model. If the number of the effects is huge, bulked segregant analysis were used to test which effects were associated with the trait. All the selected effects were further shrunk by empirical Bayesian, so significant effects could be identified. A series of Monte Carlo simulations was performed to validate the new method. Furthermore, all the significant effects were used to calculate genotypic values of all the missing F1 hybrids, and all these F1 phenotypic or genotypic values were used to predict elite parents and parental combinations. Finally, the new method was adopted to dissect the genetic foundation of oil content in 441 rapeseed parents and 284 F1 hybrids. As a result, 8 main-effect QTL and 37 interacted QTL were found and used to predict 10 elite restorer lines, 10 elite sterile lines and 10 elite parental crosses. Similar results across various methods and in previous studies and a high correlation coefficient (0.76) between the predicted and observed phenotypes validated the proposed method in this study.

show abstract

Section: Methodsmentioning

confidence: 99%

Interacted QTL Mapping in Partial NCII Design Provides Evidences for Breeding by Design

Zhao

et al. 2015

PLoS ONE

View full text Add to dashboard Cite

show abstract

“…model selection and shrinkage estimation. The latter includes the fully Bayesian [37], the penalized maximum likelihood [38,39], least absolute shrinkage and selection operator, and empirical Bayes (E-Bayes) [32,40]. The fully Bayesian method should be optimal in theory because it captures the most information from the data, but it is computationally intensive; the penalized maximum likelihood is an approximation to the fully Bayesian approach, while its main advantages over other methods are reflected by its simplicity and fast speed; least absolute shrinkage and selection operator is a good approach but the shrinkage is too weak for spurious effects and too strong for large effects; and the E-Bayes provides the optimal estimates of variance components, because shrinkage is very selective, with large effects subject to virtually no shrinkage while small effects are shrunk to zero [40].…”

Section: Genetic Parameter Estimationmentioning

confidence: 99%

Genome-wide mapping of QTL associated with heterosis in the RIL-based NCIII design

Zhang

2012

Chin. Sci. Bull.

View full text Add to dashboard Cite

Heterosis represents one of the most revolutionary advancements in crop improvement. In the genetic dissection of heterosis, NCIII design is one of the most powerful and widely used mating schemes. However, the methodologies for quantitative trait loci (QTL) detection in the design were mostly based on composite interval mapping. Therefore, in this study, our purpose was to develop a statistical method for mapping epistatic QTL associated with heterosis in the RIL-based NCIII design. First, we derivated the expectations of two classical linear transformations, Z 1 and Z 2 , while a quantitative trait was controlled by two QTL with digenic epistasis and arbitrary linkage under the F ∞ and F 2 metric models. Then, we constructed an epistatic genetic model that includes all markers on the whole genome simultaneously, and estimated all the parameters in the model by the empirical Bayes approach. Finally, a series of Monte Carlo simulation experiments was carried out to confirm the proposed approach. The results show that: (1) all the augmented genetic parameters for main-effect QTL could be rightly identified with satisfactory statistical power and precision; (2) the statistical powers in the detection of augmented epistatic effects were substantively affected by the signs of pure epistatic effects; (3) it is more difficult to detect epistatic QTL than to detect main-effect QTL; (4) statistical power is higher in the RIL-based NCIII design than in the F 2 -based NCIII design, especially in the detection of the augmented epistatic effect that consists of two pure epistatic effects in opposite directions.empirical Bayes, epistasis, NCIII design, heterosis, quantitative trait locus, recombinant inbred line Citation:He X H, Hu Z L, Zhang Y M. Genome-wide mapping of QTL associated with heterosis in the RIL-based NCIII design.

show abstract

“…Although the current single variant methods of GWAS have succeeded in identifying QTNs associated with the interested traits, these approaches fail to consider the joint minor effect of multiple genetic markers on a trait ( Tamba et al, 2017 ); furthermore, they do not match the internal genetic mechanism of these quantitative traits ( Tamba et al, 2017 ; Zhang et al, 2017 ; Sun et al, 2019 ; Wen et al, 2019 ). To overcome this drawback, multi-locus methodologies have been developed, such as least absolute shrinkage and selection operator (lasso) ( Tibshirani, 1996 ; Xu, 2010 ; Zhang et al, 2012 ), Bayesian lasso ( Yi and Xu, 2008 ), adaptive mixed lasso ( Wang et al, 2011 ), and empirical Bayes ( Xu, 2007 ). All SNPs can be included in the model and can be simultaneously estimated by using multi-locus methodologies.…”

Section: Introductionmentioning

confidence: 99%

A Fast Multi-Locus Ridge Regression Algorithm for High-Dimensional Genome-Wide Association Studies

et al. 2021

View full text Add to dashboard Cite

The mixed linear model (MLM) has been widely used in genome-wide association study (GWAS) to dissect quantitative traits in human, animal, and plant genetics. Most methodologies consider all single nucleotide polymorphism (SNP) effects as random effects under the MLM framework, which fail to detect the joint minor effect of multiple genetic markers on a trait. Therefore, polygenes with minor effects remain largely unexplored in today’s big data era. In this study, we developed a new algorithm under the MLM framework, which is called the fast multi-locus ridge regression (FastRR) algorithm. The FastRR algorithm first whitens the covariance matrix of the polygenic matrix K and environmental noise, then selects potentially related SNPs among large scale markers, which have a high correlation with the target trait, and finally analyzes the subset variables using a multi-locus deshrinking ridge regression for true quantitative trait nucleotide (QTN) detection. Results from the analyses of both simulated and real data show that the FastRR algorithm is more powerful for both large and small QTN detection, more accurate in QTN effect estimation, and has more stable results under various polygenic backgrounds. Moreover, compared with existing methods, the FastRR algorithm has the advantage of high computing speed. In conclusion, the FastRR algorithm provides an alternative algorithm for multi-locus GWAS in high dimensional genomic datasets.

show abstract

Bias correction for estimated QTL effects using the penalized maximum likelihood method

Cited by 10 publications

References 35 publications

Interacted QTL Mapping in Partial NCII Design Provides Evidences for Breeding by Design

Interacted QTL Mapping in Partial NCII Design Provides Evidences for Breeding by Design

Genome-wide mapping of QTL associated with heterosis in the RIL-based NCIII design

A Fast Multi-Locus Ridge Regression Algorithm for High-Dimensional Genome-Wide Association Studies

Contact Info

Product

Resources

About