Avg_GRM_self • TSDFGS

The mean relationship within the training set; minimizing it spreads the training individuals apart.

Type	Relationship
Direction	Smaller is better
Uses a test set	no (a test set given to `optTrain()` is ignored)
Score a given set	`avg_grm_self_score()`
Search with it	`optTrain(method = "AvgGRMself")`

Definition

$\mathbf{1}_n' G_{TT}\mathbf{1}_n/n^2$ with $G = XX'/p$ .

Use in TSDFGS

library(TSDFGS)
data(geno)
avg_grm_self_score(geno[1:50, ])

## [1] 1.412919

fit <- optTrain(geno, cand = 1:404, n.train = 50, method = "AvgGRMself",
                 min.iter = 300, console = FALSE, seed = 1)
fit$OPTtrain

##  [1] 123 351 129  97 137 187 350  85 143 189 119 104 389 176 160 346 272 217 393
## [20] 127 290 288 382  89 289   1 360 331  42 125  20  44 175  70  30  40 163  25
## [39] 259  17  39 298 111  76  14 130  45 196 281 230

The search is a random-start exchange (hill-climbing) search; see How the scores are computed. Use min.iter larger than 300 on real data and check fit$converged.

Original work

Fernández-González, J., Akdemir, D., & Isidro y Sánchez, J. (2023). A comparison of methods for training population optimization in genomic selection. Theoretical and Applied Genetics, 136, 30. https://doi.org/10.1007/s00122-023-04265-6

Repository or website of the original authors: Code of Fernández-González et al. (2023): github.com/TheRocinante-lab/Publications. Package TrainSel: github.com/TheRocinante-lab/TrainSel.

License review

Formulas and algorithms are ideas and methods, which copyright does not protect. Where an original package is under a GPL license, we read it to check our results and credit it; TSDFGS is GPL-3, which is compatible with GPL-3 and GPL (>= 3) code. This is our reading of the public record, not legal advice; see the license review.

Original software: the repository of Fernández-González et al. (2023) has no LICENSE file, and TrainSel carries an all-rights-reserved license with a patent notice. We did not read or copy either; the definition comes from the open-access article.
Verdict: included, written from the article.

Reviewed 2026-11-04. All criteria and the ones left out.