Skip to contents

CDmean

Source: vignettes/articles/criterion-cdmean.Rmd
criterion-cdmean.Rmd

The mean coefficient of determination of the GBLUP contrast between each target individual and the target mean (Rincent et al., 2012).

Type Model-based
Direction Larger is better
Uses a test set yes; without a test set, all candidates
Score a given set cdmean_score()
Search with it optTrain(method = "CDmean")

Definition

With K=XX′/pK = XX'/p, M=I−J/nM = I - J/n and W=(MKT+λI)−1MW = (MK_T + \lambda I)^{-1}M (λ=1\lambda = 1): for each target ii with contrast ci=ei−𝟏/nPc_i = e_i - \mathbf{1}/n_P, CDi=vi′Wvi/(ci′Kci)\mathrm{CD}_i = v_i'Wv_i/(c_i'Kc_i) with vi=KTPciv_i = K_{TP}c_i; the score is the mean. Equal to the CDMEAN of STPGA in our tests.

Use in TSDFGS 

library(TSDFGS)
data(geno)
cdmean_score(geno[1:50, ], geno[51:100, ])
## [1] 0.6039068
fit <- optTrain(geno, cand = 1:404, n.train = 50, method = "CDmean",
                 min.iter = 300, console = FALSE, seed = 1)
fit$OPTtrain
##  [1] 181 312 119 184 178 170 251 256 277   5 330 263  32  79 158 135 105 217 199
## [20]  73 391 287 373  89 289 370 153 394  42 237  20  44 308  92 121 247 205 389
## [39] 262 221  62 298 165 160  38  15 183 196 206 230
# targeted: the test set is known
fit2 <- optTrain(geno, cand = 1:300, n.train = 50, test = 301:404,
                  method = "CDmean", min.iter = 300, console = FALSE, seed = 1)
fit2$OPTtrain
##  [1] 126 129  76   1  85  55 264 293 298  43  74 217 165  89 278  67  35  44 281
## [20]  34  40 216 280 225 289 226 160 220  65 133 245 283 253 262 300  71 202   9
## [39]  13  17  31 124 260  99 240 151   4 162 268 122

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

Rincent, R., Laloë, D., Nicolas, S., Altmann, T., Brunel, D., Revilla, P., et al. (2012). Maximizing the reliability of genomic selection by optimizing the calibration set of reference individuals: comparison of methods in two diverse groups of maize inbreds (Zea mays L.). Genetics, 192, 715-728. https://doi.org/10.1534/genetics.112.141473

Chen, S.-P., Sung, W.-H., & Liao, C.-T. (2024). Constructing training sets for genomic selection to identify superior genotypes in candidate populations. Theoretical and Applied Genetics, 137, 270. https://doi.org/10.1007/s00122-024-04766-y

Repository or website of the original authors: R package STPGA by Akdemir and colleagues: CRAN.R-project.org/package=STPGA (GPL-3). R package TSODBG of Chen et al. (2024): github.com/spcspin/TSODBG (GPL >= 3).

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: STPGA and TSODBG, GPL; compared, not copied.
  • Article license: Rincent et al. (2012) CHORUS.
  • Verdict: included.

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