Variable-Selection Emerges on Top in Empirical Comparison of Whole-Genome Complex-Trait Prediction Methods

Haws, David; Rish, Irina; Teyssèdre, Simon; He, Dan; Lozano, Aurélie; Kambadur, Prabhanjan; Karaman, Zivan; Parida, Laxmi

doi:10.1371/journal.pone.0138903

Cited by 18 publications

(23 citation statements)

References 53 publications

(70 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the remainder of the paper we will focus on the elastic net, which we have found to outperform other predictive models on real-world data [31]. This has been recently confirmed in [32]. …”

Section: Methodsmentioning

confidence: 90%

Using Genetic Distance to Infer the Accuracy of Genomic Prediction

2016

View full text Add to dashboard Cite

The prediction of phenotypic traits using high-density genomic data has many applications such as the selection of plants and animals of commercial interest; and it is expected to play an increasing role in medical diagnostics. Statistical models used for this task are usually tested using cross-validation, which implicitly assumes that new individuals (whose phenotypes we would like to predict) originate from the same population the genomic prediction model is trained on. In this paper we propose an approach based on clustering and resampling to investigate the effect of increasing genetic distance between training and target populations when predicting quantitative traits. This is important for plant and animal genetics, where genomic selection programs rely on the precision of predictions in future rounds of breeding. Therefore, estimating how quickly predictive accuracy decays is important in deciding which training population to use and how often the model has to be recalibrated. We find that the correlation between true and predicted values decays approximately linearly with respect to either FST or mean kinship between the training and the target populations. We illustrate this relationship using simulations and a collection of data sets from mice, wheat and human genetics.

show abstract

Section: Methodsmentioning

confidence: 90%

Using Genetic Distance to Infer the Accuracy of Genomic Prediction

2016

View full text Add to dashboard Cite

show abstract

“…The identification of markers related to this phenotype using FS is based on these techniques to provide an interpretable model due to the close relation between trait and genotype; i.e., using the subset of high-density markers might help elucidate the regions most likely to be involved in phenotypic differentiation 120 . This strategy of selecting a subgroup of SNPs with higher predictive power and closeness to the predictive class has already been employed in different contexts 48,121,122 .…”

Section: Discussionmentioning

confidence: 99%

Machine learning approaches reveal genomic regions associated with sugarcane brown rust resistance

Aono

Costa

Rody

et al. 2020

Preprint

View full text Add to dashboard Cite

Sugarcane is an economically important crop, but its genomic complexity has hindered advances in molecular approaches for genetic breeding. New cultivars are released based on the identification of interesting traits, and for sugarcane, brown rust resistance is a desirable characteristic due to the large economic impact of the disease. Although marker-assisted selection for rust resistance has been successful, the genes involved are still unknown, and the associated regions vary among cultivars, thus restricting methodological generalization. We used genotyping by sequencing of full-sib progeny to relate genomic regions with brown rust phenotypes. We established a pipeline to identify reliable SNPs in complex polyploid data, which were used for phenotypic prediction via machine learning. We identified 14,540 SNPs, which led to a mean prediction accuracy of 50% by using different models. We also tested feature selection algorithms to increase predictive accuracy, resulting in a reduced dataset with more explanatory power for rust phenotypes. Using different feature selection techniques, we achieved accuracy of up to 95% with a dataset of 131 SNPs related to brown rust QTL regions and auxiliary genes. Therefore, our novel strategy has the potential to assist studies of the genomic organization of brown rust resistance in sugarcane. 3/154/15 10/15

show abstract

“…The agreement between the measures of marginal effect suggests that the distribution of marginal effects is representative. Haws et al (2015) worked on a problem similar to ours in the area of genomic selection, i.e. very many independent variables and relatively few observations.…”

Section: Discussionmentioning

confidence: 99%

“…very many independent variables and relatively few observations. The study of Haws et al (2015) highlights the influence that the choice of analysis method has on the conclusions that can be made, which is why we took care to triangulate our estimates of marginal effect. It must be noted that our conclusions are limited to the dataset, metrics and statistics used.…”

Section: Discussionmentioning

confidence: 99%

“…If there are an obvious subset of inputs that strongly relate to the output then univariate selection methods work best. Haws et al (2015) suggested that this could be evidenced by having a small number of strong marginal effects. Examples of univariate variable-selection methods include measures of marginal effects or interaction effect methods like minimum-redundency-maximal-relevance (Peng et al, 2005; see Ng and Breiman, 2005 for discussion on univariate and bivariate variable selection methods).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Spatio-temporal metrics that distinguish play outcomes: A pilot study

McInerney

Goodwill

Foster

et al. 2017

jhse

View full text Add to dashboard Cite

In team invasion sports, tactical behaviour can be examined using spatio-temporal data, i.e. the position of the players at a given time. A review of the spatio-temporal metrics used in team invasion sports performance analysis indicated that thousands of variations of metrics being used. Information about the distribution of metrics' individual effects can inform us of the best variable-selection method. The aim of this pilot study was to estimate the distribution of strong marginal effects of spatio-temporal metrics of field hockey plays. With institutional ethical approval, the Womens' and Mens' gold medal games from the EuroHockey 2015 field hockey tournament were recorded. Best, acceptable and worst-case outcomes for plays were described by 1,837 spatio-temporal metrics. Each metric's marginal effects were estimated using Cramér's V, Mutual Information and the I-score. Values for Cramér's V of 0.2 and 0.4 to mark the boundaries of small, moderate and large effects. Less than 1% of metrics show large effects with > 87% of all metrics showing small effects as per the Cramér's V thresholds. These large effect metrics where all within the 98th percentile of Mutual Information values and within the 96th percentile of the I-score values, which supports the Cramér's V distribution of marginal effects. Therefore, according to the recommendations of Tibshirani (1996), univariate variable-selection methods will be the most appropriate for selecting important metrics.

show abstract

Variable-Selection Emerges on Top in Empirical Comparison of Whole-Genome Complex-Trait Prediction Methods

Cited by 18 publications

References 53 publications

Using Genetic Distance to Infer the Accuracy of Genomic Prediction

Using Genetic Distance to Infer the Accuracy of Genomic Prediction

Machine learning approaches reveal genomic regions associated with sugarcane brown rust resistance

Spatio-temporal metrics that distinguish play outcomes: A pilot study

Contact Info

Product

Resources

About