Shulou(Shulou.com)06/01 Report--

This article introduces whether to use blue value or blue value in GWAS and GS analysis. The content is very detailed. Interested friends can use it for reference. I hope it will be helpful to you.

Question:

In plants, there are multiple phenotypic values for each genotype based on multi-year or multi-year data. The question is, if you need a genotype corresponding to a phenotypic value when doing GWAS or GS, what is the phenotypic value?

At present, there are three answers:

1, average

2, BLUE value (best linear unbiased estimation, fixed factor)

3, BLUP value (best linear unbiased prediction, random factor)

Discussion in the forum:

Https://www.researchgate.net/post/Does_any_one_have_an_idea_of_which_one_BLUE_or_BLUP_to_use_for_a_GWAS_analysis_of_a_trait_in_wheat_eg_resistance_to_rust

Does anyone know which blue or BLUP to use for GWAS analysis of wheat traits (for example, rust resistance), I have three resistance assessment data sets from two sites generated in field trials designed by Alpha lattice (2 repeats of 300 materials, each repeat, 10 incomplete blocks, containing 30 materials).

Two of the datasets come from the same location but in different years; the third is one-year data from another location).

Therefore, I am considering calculating the Blu-ray / Blu-ray at each location and calculating the total Blu-ray / Blu-ray for the combined data used in GWAS.

Answer 1:

I strongly recommend using blue because you are doing a two-stage analysis.

Blue will allow you to provide an "adjusted average" for each genotype based on the design effect and other covariates in the model.

This is what you want, a more precise approach. This translates into using your model effects (for example, randomly copying fixed and incomplete blocks and graphs), but your genotype (or clone) effects are fixed.

If you use BLUP, you are shrinking your genetic effects.

This will mean that your genetic effects will move to the average based on this information, yes, they will be your best predictions of these random effects, but they will be adjusted according to the sample size and the variance associated with these data.

This is the effect of random effects, but the problem is that if it is random, you will eliminate some of the genetic signals and then tend to get more noise than you want to GWAS.

Therefore, your fitness is fixed, once you have done your GWAS, it will be a random effect, but it will not double shrink.

Good luck.

Answer 2:

Hi, Sisai.

I suggest that all design effects are random (the attribute is larger, not the effect), and the overall structure and tag effect are fixed (the attribute is blue, not the effect).

There is no reason to analyze it separately in each test.

One-stage analysis is always desirable (all model parameters are learned from the same possibility).

If you use the above approach, there may be several risks associated with false positives.

If you don't have confidence in one-stage analysis, I suggest you analyze each trial individually and treat all design effects as random and genotypic effects fixed.

In GWAS analysis, marker effects are modeled as fixed effects, and relational matrices (including genotypic random effects in the model) are used to control the population structure, or the population structure is modeled as a fixed effect, or both.

You can use unsupervised or supervised clustering methods (PCA, structure, and so on) to identify subgroups.

I wish I could help.

Overall conclusion:

1. If you can use one-stage, that is, the combined analysis of one-year multi-point and multi-year multi-point data, instead of calculating the correction value first and then analyzing it (that is two-stage), it is best not to use any value as the phenotypic value, but to use the original phenotypic value directly.

2. If the two-stage must be carried out, that is, the correction value must be calculated first, and then GS or GWAS, it is recommended to use the blue value of the variety instead of the blue value. Because in the mixed linear model, the random factor will shrink to the mean (shrinkage), although the result is the best prediction, but the variance of the correction value becomes smaller, when you do GWAS, it is not easy to find significant sites, increasing the noise (noise). And in GWAS or GS, the variety is used as a random factor, and if you use the blup value, it is equivalent to two shrinkage.

3, therefore, a better way is to calculate the blue value in one-stage by taking place, year and block as random factors and varieties as fixed factors.

Reference 1:

This article is an article of winter wheat GS.

Here, the plot heritability is calculated and the residual is divided by the harmonic average of each genotype. At the same time, in the model, the variety was taken as a fixed factor, and the blue value was calculated.

Here, the blue value is used instead of the BLUP value in animal breeding.

The heritability of 13 environments was estimated and the blue value of each environment was calculated.

In wheat breeding, the key selection is the genotype value, not the breeding value, so the BLUE value is more suitable than the blup value. (this sentence is a little confusing)

Reference 2:

Https://www.researchgate.net/publication/268118579_Genomic_Selection_for_End-Use_Quality_Traits_in_CIMMYT_Spring_Wheat

In reference 2, LSmeans is taken as blue value.

On GWAS and GS analysis in the end is to use blue value or blup value to share here, I hope the above content can be of some help to everyone, can learn more knowledge. If you think the article is good, you can share it for more people to see.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.