Using partial least squares regression to determine variable importance in expected genetic gain in open nucleus breeding scheme to improve milk production in Egyptian buffalo

الملخص EN

The purpose of this paper was to study variable importance in the projection (VIP) determining the expected genetic gain when an applying open nucleus breeding scheme (ONBs) to improve milk production of Egyptian buffaloes.

Populations of buffaloes with one milk record for each cow were generated with assumed mean (0) and variance (1).

Four different population sizes (z) : 10000, 25000, 50000 and 100000 cows were generated.

Four generations of progeny were obtained by selection of sires and dams of the next generations.

The structure of ONBs was applied for different population size.

Genetic gain per generation was calculated for the whole system of ONBs (GGS), base population (GGB) and nucleus population (GGN).

A high correlation was observed between GGS and generation (G), mating ratio (MR) and fraction of base sires born in nucleus (w).

The high correlation between explanatory variables was observed between fraction of total population in nucleus (p) and fraction of base dams born in nucleus (y).

G, MR and w had high influence in the model (VIP > 1).

The goodness of fit coefficients, the analysis of the model corresponding to GGS allow to conclude that the model is well fitted (R2 = 0.77).

The G had the highest weight in the model of GGS.

Under cross-validation, the model of predicted GGS was fit.

MR had the highest influence on GGB and that reflect the importance of high selection intensity in maximizing the genetic gain in base population.

The G and MR had high influence in the model (VIP > 1) of GGN.

While, w had a moderate (0.8 < VIP < 1) influence on GGN.