Building the Embrapa rice breeding dataset for efficient data reuse

Embrapa has led breeding programs for irrigated and upland rice (Oryza sativa L.) since 1977, generating a large amount of pedigree and phenotypic data. However, there were no systematic standards for data recording nor long‐term data preservation and reuse strategies. With the new aim of making data reuse practical, we recovered all data available and structured it into the Embrapa Rice Breeding Dataset (ERBD). In its current version, the ERBD includes 20,504 crosses involving 9,974 parents, the pedigrees of most of the 4,532 inbred lines that took part in advanced field trials, and phenotypic data from 2,711 field trials (1,118 irrigated, 1,593 upland trials), representing 226,458 field plots. Those trials were conducted over 38 years (1982–2019), in 247 locations, in latitudes ranging from 3°N to 33°S. Phenotypic traits included grain yield, days to flowering, plant height, canopy lodging, and five important fungal diseases: leaf blast, panicle blast, brown spot, leaf scald, and grain discoloration. The total number of data points surpasses 1.27 million. Descriptive statistics were computed over the dataset, split by cropping systems (irrigated or upland). The mean heritability of grain yield was high for both systems, at around .7, whereas the mean coefficient of variation was 13.9% for irrigated trials and 18.7% for upland trials. The ERBD offers the possibility of conducting studies on different aspects of rice breeding and genetics, including genetic gain, G×E analysis, genome‐wide association studies and genomic prediction. Core Ideas Historical datasets are useful for genetic research and development of breeding tools. Phenotypic data from crop breeding programs connect elite genotypes with target environments. Converting available records in analyzable format is a worthwhile investment.