Comparative study of univariate and multivariate selection strategies based on an integrated approach applied to oilseed rape breeding

Abstract

AbstractChoosing the right method to discover superior genotypes is always one of the major challenges that oilseed rape (Brassica napus L.) researchers deal with. This work aimed to compare the univariate and multivariate selection indices utilizing variance‐based, additive main effects and multiplicative interaction based, best linear unbiased prediction based, and genotype × environment/trait based strategies to determine the best‐performing (regarding stability, yield, and agronomic traits) oilseed rape genotypes. Utilized indices were categorized into five general methods, including univariate performance selection index (UPSI: mean performance), univariate stability selection indices (USSIs: environmental variance, Shukla's stability variance, AMMI stability value, modified AMMI stability value, modified AMMI stability index, and weighted average of absolute scores), multivariate performance selection indices (MPSIs: genotype by trait [GT], genotype by yield × trait [GYT], and ideotype‐design index [FAI‐BLUP]), univariate performance and stability selection indices (UPSSIs: coefficient of variability, harmonic mean of genotypic value [HMGV], relative performance of genotypic value [RPGV], WAASBY [WAASB and mean performance {Y}], and genotype + genotype × environment interaction [GGE]), and multivariate performance and stability selection index (MPSSI: MTSI). Data from 18 oilseed rape genotypes and two checks were obtained from trials conducted at eight locations in Iran during two consecutive crop years from 2018 to 2020. The grain yield (GY) and agronomic traits were evaluated. The significant genotype × environment interaction was found for GY and agronomic traits in the combined analysis of variance. In terms of GY, the selection indices GGE, HMGV, and RPGV (6.51), GYT (6.43), GT (5.35), MTSI (4.95), FAI‐BLUP (4.64), and WAASBY (3.97) achieved better‐than‐average genetic gains (GSs). Using an optimized model of USSIs (O‐USSIs), GS increased from −0.15 to 3.26 on average, which shows the efficiency of O‐USSIs in high‐yielding and stable genotypes selection. Based on Spearman's ranking correlation, a positive relationship was found among MPSIs, MPSSI, UPSSIs, and UPSIs, but not for USSIs. Based on an integrated model of selection indices, genotypes G2, G10, G11, G14, and G20 were detected as superior genotypes with high‐yielding, high‐stability, and desirable agronomic traits. It could be concluded that combining different approaches of genotype selection based on an integrated model can provide better answers to oilseed rape breeders.