To select the optimal subset of variables (descriptors) that can significantly correlate with biological activity of molecules from the pool of descriptors, various variable selection methods such as step-wise search algorithm, genetic algorithm, and simulated annealing among others can be used.
Other variable screening methods such as expert opinions, principal component analysis (PCA), cluster analysis, and partial least squares (PLS) are also applied to compare the performance of the proposed method.
Furthermore, efficient variable selection methods, such as least absolute shrinkage and selection operator, least angle regression and forward selection, have been effectively incorporated into the improved model for the determination of regression coefficients.
These limitations may be addressed with other variable selection methods such as model averaging [ 19].
Results confirmed our assumption, and BLUP|GA outperformed GBLUP/RRBLUP and even variable selection methods such as BayesA.
Stability selection is a general framework to combine variable selection methods such as penalized regression models with subsampling strategies.
This is commonly done, for example, in Bayesian LASSO and stochastic search variable selection methods such as BayesCn and BayesDn (see e.g. [ 52]).
For this particular case, it is shown in several studies that models with a thick-tailed prior distribution of marker effects such as BayesA and variable selection methods such as Bayes SSVS yield higher accuracy than G-BLUP.
These range from variable selection methods such as BayesB, which allows only a small number of loci to have an effect, some of them potentially large, to gBLUP, which assumes equal variance across all loci [ 9].
