GS4PB: An R Shiny application to facilitate a genomic selection pipeline for plant breeding

The implementation of genomics-assisted breeding methodologies is helping to drive the genetic gain required to meet the grand challenge of producing more food using fewer resources in the face of a changing climate. Despite the documented usefulness of genomics-assisted breeding toward this end, its full infusion into most small- and medium-sized breeding programs is still incomplete. One major reason for limited routine application of genomic selection among most such programs is the lack of a single integrated software tool capable of assisting breeders throughout the entire genomic prediction pipeline. To help address this need, we have implemented a streamlined genomic prediction and selection pipeline designed for plant breeding programs using open-source tools. The steps implemented in the pipeline include processing genotypic data (e.g., filtering and imputing genotypic data), merging genotypic and phenotypic data, collecting enviromics covariates, estimating environmental kinship, optimizing training sets, cross-validating genomic prediction models, and implementing genomic prediction for single or multiple traits across single or multiple environments. Herein, we describe an R Shiny web application named "GS4PB" (Genomic Selection For Plant Breeding) that implements the above steps in the pipeline and discuss the rationale for each of the tools in the pipeline. We used this GS4PB application to conduct an experiment comparing phenotypic and genomic selection, and showed genomic selection worked as well as phenotypic selection for advancement of breeding lines. This publicly available analysis tool will help to lower entry barriers into advanced techniques of genomic prediction, enabling breeders to take advantage of these technologies to help drive genetic gain.