Objective

G2F is a multi-institutional, collaborative initiative to develop tools that efficiently predict performance of diverse maize (Zea mays ssp. mays) varieties across multiple growing conditions. G2F projects aim to collect, share, and analyze multi-year, large-scale genomic, phenotypic, and environmental datasets. The project builds on existing maize genome sequence resources by developing approaches to understand the functions of genes and specific alleles based on their expression in typical field conditions. There are many dimensions to the goal of understanding genotype-by-environment (G × E) interactions, including which genes impact which traits and trait components, how genes interact among themselves, the relevance of specific genes under different growing conditions, and how genes influence plant growth during various stages of development.

G2F projects foster integration of diverse research disciplines, including (but not limited to) genetics, genomics, plant physiology, agronomy, climatology, and crop modeling as well as analytical perspectives and tools derived from computational sciences, statistics, and engineering. Under the umbrella of G2F are enterprises such as the G × E project that began in 2014. The G × E project aims to document and measure genotypes, phenotypes, and environmental data in standard formats across more than twenty distributed field locations in North America annually. The resulting dataset is unique as it represents, to our knowledge, the most extensive publicly available dataset of its kind, reporting a consistent set of traits across common sets of fully genotyped germplasm not only across many locations, but also with relevant information reported down to the level of specific plots. Making these datasets publicly available enables researchers from many different disciplines to tackle the daunting analyses necessary to make useful predictions of crop performance. Novel data analysis approaches and tools are expected to result from the curated and organized data described here.

Data description

Online forms were developed for logging field site coordinates, field management metadata, and other site-specific information. Datasets include:

  • DNA sequences of inbreds (with and without imputation), including those inbreds used to produce featured hybrids. The process for creating files and metadata pertaining to the genotype by sequencing (GBS) process [1] is described. Data are most readily analyzed using TASSEL software [2]. Raw sequence reads generated are accessible via the Sequence Read Archive [3].

  • Phenotype measurements for inbreds and hybrids. A handbook of instructions for making traditional phenotype measurements (reviewed in [4]) is available via the G2F website [5]. Traditional traits include stand count, stalk lodging, root lodging, days to anthesis, days to silking, ear height, plant height, plot weight, grain moisture, and test weight. Datatypes reported as both raw files and files with outliers removed are described in README files. Additionally, a large set of ear, cob, and kernel measurements was made with a non-traditional machine vision platform to quantify the components of yield [6]. These data are reported in millimeters with shape descriptors reported as principal components of contour data points. Cob color was reported as RGB (red/green/blue) pixel values. Kernel row number, counted manually, is reported as an integer.

  • Environmental data collected by WatchDog 2700 weather stations (Spectrum Technologies) at 30-min intervals from planting through harvest. Collected information includes wind speed, direction, and gust; air temperature, dewpoint, and relative humidity; rainfall; and solar radiation. Data are reported as a calibrated set (based on calibration derived from nearby National Weather Service stations) and “clean” (based on removing obvious artifacts from the calibrated dataset).

  • Soil characterizations by site (first taken in 2015) including plow depth, pH, buffered pH, organic matter, phosphorus levels (in parts per million), and potassium levels (in parts per million).

Data collected in year n are released to project members in spring of the following year (n + 1), and released to the public the subsequent year (n + 2). The 2014 and 2015 datasets are publicly available via the NCBI SRA [7] and CyVerse/iPlant [8] with files and access links shown in Table 1.

Table 1 Overview of data files and data sets
Full size table

As technologies develop and the number of researchers involved in the project grows, it is anticipated that increasingly diverse datatypes will be documented. An example of the use of these data has been reported [12]. In that study, phenotypic plasticity was found to be disproportionately controlled by regulatory regions. Because these datasets support lines of inquiry limited only by the questions researchers pose, the potential scope of application for these data is broad. The dataset is anticipated to additionally impact the field simply by being the first public dataset of its scale that has been collected and reported using standardized protocols and formats, respectively, thus defining standards for data collection, formatting, and access.

Limitations

Missing data occurs in most datasets. For genotypic and phenotypic datasets, missing data are left blank rather than zero or ‘null’ representation because some measured data report zero values and some software will only accept numeric values (not strings). The exception is for traits extracted from inbred ear, cob, and kernel image data, which are demarcated with ‘NA’.

In some instances, reported data were maintained rather than editing for consistency. These decisions were made to minimize misinterpretation that could lead to incorrect documentation or measurements.

For weather data, raw files reported by sensors are not provided because machine data were calibrated based on information from nearby weather stations to ensure accuracy (e.g., if the wind vane was set improperly, a calibration correction was required).

Field locations are not always identical year-to-year, primarily due to crop rotation management practices. Each field’s GPS coordinates are reported annually to enable data aggregation in keeping with specific research objectives.

Germplasm used and reported are specific to the project and are held by researchers involved in the project. They do not derive directly from national public genebanks. Seed access is granted in keeping with seed availability from cooperating researchers directly.