ALLELIC COMBINATIONS OF VRN, PPD ED RHT GENES IMPROVE GENOME ENABLED PREDICTION MODELS IN DURUM WHEAT
Author(s)
Date Issued
2024
Type
conferenceObject
Abstract
The presence of different genetic variations (e.g. SNP, INDEL) at
vernalization (Vrn), photoperiod (Ppd), and plant height (Rht) genes play a
key role in the success of environmental adaptation of durum wheat.
Studying their diversity, frequency, and interactions will aid in
determining which allele combinations (ACs) are effective and beneficial
for a specific growing condition to maximize yields, particularly in the
face of climate change. In this context, genomic prediction (GP) is a
useful breeding-tool as it can support crop breeding programs exploiting
allelic diversity. Hence, the focus of this research was to test the effect
of AC on the predictive ability of GP models in a panel of 190 durum wheat
genotypes, including landraces, old and modern cultivars. The genetic
material was genotyped with the Illumina iSelect 15K SNPs assay, whereas
diagnostic molecular markers at Vrn-1, Vrn-3, Ppd-1 and Rht-1 loci were
used for defining ACs. Individuals were phenotyped for heading date (HD),
flowering time (FT), and plant height (PH) across three different sowing
dates (early, optimal, and delayed) over three consecutive growing seasons
at CREA, Foggia (IT). RKHS regression GP models were fitted with a linear
kernel method (GB) and the non-linear Gaussian Kernel (GK) method. In
addition, for each GB and GK method, the ACs were included as fixed effects
(GB-AC and GK-AC), and two other GP models were implemented. A leave-one
out and k-fold cross-validation strategies were used to estimate the
predictive ability (PA) of GP models using Pearson correlation coefficient.
Three alleles were found for Vrn-A1 and Ppd-A1, two for Ppd-B1 and Rht-B1,
while Vrn-B1 and Vrn-B3 were monomorphic, with the recessive allele present
in all samples. Dominant alleles such as Vrn-A1c, Ppd-A1b, and Rht-B1b
where abounded within modern cultivars, accounting for 77.2%, 65.3% and 97%
respectively, highlighting the impact of breeding on pheno-morphological
traits. Twenty-two ACs were identified in the entire panel, with AC09
showing the highest frequency among landraces (41.9%) and old cultivars
(30.4%), and AC04 the highest among modern cultivars (32.7%). Single
alleles and ACs significantly impacted the duration of the phenological
phases. In particular, AC2 and AC16 anticipated HD and FT, whereas AC7 and
AC21 delayed plant development.
Comparing the non-linear GK with the conventional linear GB method, an
average increase in PA ranging from 0.20 to 0.22 for FT and from 0.47 to
0.52 for PH was obtained using GB. Interestingly, integrating the allelic
combinations into both GP models significantly improved the model accuracy
for all traits and sowing-by-season combinations. For instance, using GB-AC
method PA ranked from 0.22 to 0.50 for FT, and from 0.52 to 0.74 for PH,
suggesting that this method might be used to predict pheno-morphological
traits, paving a practical way to support durum wheat breeding programs.
Conference(s)
LXVII SIGA Annual Congress