Selection Strategy For The Beverage Sensory Characterization In A Large Arabica Coffee Germplasm Bank

40 This study proposes a selection strategy to be applied to a large number of coffee accessions, 41 which can be useful when exploring many genotypes in breeding programs for beverage quality. Over 42 three consecutive years, 270 Arabica coffee accessions from the Germplasm Bank of Minas Gerais, 43 Brazil, were sensorially characterized. At the end of this period, the 20 genotypes with the greatest 44 potential for specialty coffee production were selected, which were subjected to a second sensory 45 characterization. Additionally, we determined the genetic correlations between the attributes that make up 46 the final beverage score. We did not identify an attribute that was more important than the others in the 47 organoleptic performance of the studied accessions. Through the proposed method, we identified genetic 48 variation in our original population and selected five accessions with beverage quality considered 49 excellent according to the SCA classification. These accessions were planted in field conditions of highly 50 specialized farms to study their interaction with the cultivation environment and to promote the increase 51 in production of such coffees. Additionally, the five selected accessions are being used in crosses with 52 elite cultivars create new segregating populations. We conclude that estimation of the heritable variation 53 of each accession, the use over the years of common check cultivars with known performance, and 54 extremely care of experimental precision in the entire processes warrant fair comparisons.

which leads to a greater demand for these varieties by the buyers of these coffees (Illy and Viani 2005).

67
However, little is known about the genetic influence of sensory traits. To fill this knowledge gap, two 68 questions need to be answered. The first question is whether beverage quality is genetically controlled, so 69 heritable. The second question is whether phenotypic superiority can be passed on to the offspring during 70 the selection process performed by the breeders. If so, it is also necessary to estimate the crop's genetic 71 progress through selection gain. The first question has been addressed by previous researchers, but with 72 other plant genotypes in the multilocation trials (Barbosa et al. 2020) and different evaluation methods 73 (Agwanda et al. 2003).

74
The evaluation of the quality of coffee is performed by physical classification and sensory 75 analysis called cupping, in which professional tasters evaluate and score various attributes of a coffee and 76 provide a description of its nuances (Lingle and Menon 2017). Some researchers use the phenotypic data 77 from this evaluation to perform the selection, considering the traits as fixed effects and without estimating 78 the genetic variance or heritability of the phenotypic variation (Silva et al. 2014;Figueiredo et al. 2015; 79 Nadaleti et al. 2018;Donfrancesco et al. 2019). However, this strategy may not be effective because this 80 procedure does not consider how much of the phenotype is due to the genotype, which does not allow us 81 to predict whether the selected superiority will hold up in the future, such as when the factors under study 82 have random effects (Smith et al. 2007;Resende et al. 2016). Understanding the genetic inheritance of 83 quality traits is essential to benefit these programs (Tran et al. 2016). Because quality traits are affected 84 by many factors, information on the association between various attributes of organoleptic quality can 85 also benefit the breeding process.
Given the above, the first objective of this study was to determine whether there is genetic 96 variability and heritability in the sensory quality of the coffee beverage. The possibility of generating 97 coffee plants with higher quality potential is extremely important, but the success of this strategy depends 98 on the existence of genetic variation in the studied trait, consequently a heritability value. The second 99 objective was to determine the genetic correlations between the attributes that make up the final beverage 100 score. This correlation will help assess whether the selection gain is indirectly greater in each of these 101 attributes. The third objective was to assess the efficiency of the selection process in a large number of 102 accessions from BAG-MG evaluated over three consecutive years. The proposed method can be useful 103 when exploring many genotypes in breeding programs. At the end of the line, the focuses are immediate 104 use of the selected accesses by producers and cross them with elite cultivars to develop segregating 105 populations.

108
The BAG-MG is located in the Alto Paranaíba region, at 18°59′26″ south latitude and 48°58′9.5″ 109 west longitude, with altitude of approximately 1,000 meters. The soil is a dystrophic red-yellow Latosol.

110
The topography is flat with a slight slope. The average annual temperature is 20.61°C, with maximum of After drying, the samples were placed in double-layer Kraft paper bags, covered with a plastic 124 bag, and stored for 30 days in a cold chamber with a constant temperature of 16 °C to standardize the 125 water content in the grains.

126
For the sensory analyses, the samples were standardized on high mesh size, and we selected the 127 grains free of intrinsic and extrinsic defects. The grains were toasted according to the protocol proposed 128 by the Specialty Coffee Association -SCA (Lingle 2011), which recommends color 55# to 65# on the 129 Agtron color scale for whole grains, in addition to a roasting time of 8-12 minutes. These analyses were 130 performed in five cups per sample by three accredited and calibrated Q-graders using the same protocol, 131 which consists of 10 sensory attributes: flavor, acidity, balance, aftertaste, body, aroma, and overall 132 (which are evaluated with scores from 6 to 10 points each) and the attributes uniformity, sweetness, and 133 clean cup (to which are assigned 2 points per cup that is free of defects, that is uniform, and that has a 134 minimum sweetness equivalent to the concentration of 0.5% w/v of sucrose). The final score was 135 calculated considering the sum of the 10 sensory attributes cited, and coffees with a final score equal to or 136 greater than 80 points were considered specialty coffees.

137
To compare the mean final score of the different accessions in the three harvests, the mean of the

142
Amarelo were used as controls in the three harvests. The data for the final sensory score of the three 143 evaluation years obtained by the estimator were subjected to statistical analysis and estimates of 144 genotypic values and heritability for each accession, using the Selegen REML/BLUP software (Resende 145 2016). The model y = Xr + Zg + e was used, where y is the data vector, r is the vector of the repetition 146 effects (assumed to be fixed) added to the overall mean, g is the vector of the genotypic effects (assumed 147 to be random), and e is the vector of (random) errors or residuals. X and Z represent the incidence The 27 accessions that had the highest predicted additive values were subjected to new sensory 153 evaluation in the 2019/2020 harvest, and their genetic parameters were estimated by the model y = Xr + 154 Zg + Wi + e, where i is the vector of the genotype × environment interaction effect (random), and W 155 represents the incidence matrix for this effect. In 2018, there was no harvest due to a negative biennial 156 cycle, so there was not enough production of plants to obtain a sample.

157
The genetic correlations between the attributes that make up the final beverage score of the 158 accessions evaluated in the 2019/2020 harvest were analyzed using the software Genes (Cruz 2013). The 159 sensory attributes uniformity, sweetness, and clean cup were not statistically evaluated, since all samples 160 were given a score of 10 on these attributes.

163
The studied population had great genotypic variability in the traits that make up beverage quality 164 (Supplemental Table S1). The predicted additive values ranged from 75.87 to 85.53, and heritability

177
The individual heritability of the selected accessions ranged from 61 to 99%, and the predicted additive 178 value ranged from 84.44 to 87.10. Considering a selection intensity of 25% (five accessions), an SG of Aiming to identify heritable associations to guide breeding programs for beverage quality, the 182 genetic parameters of the attributes that make up the final quality score and the correlations between them 183 and the final score were estimated (Table 1). The correlations between all traits studied were positive, 184 significant. Of high magnitude, ranging from 0.73 to 0.95, indicating that all attributes of organoleptic 185 quality contributed to the classification of genotypes and the final score, as expected.

186
When evaluating the genetic parameters, all the traits studied were strongly influenced by the 187 environment. The individual broad sense heritability values were low, ranging from 16 to 32%, so it can 188 be inferred that the studied accessions tend to have lower stability of expression of these characteristics in

203
There is a strong need to develop low-cost and high-yield evaluation strategies for classical 204 breeding programs, especially for perennial crops such as coffee. Basically, our study sought to clarify

243
We found correlations between all sensory attributes, corroborating other authors (Pereira et al.