Maize: GO Annotation - Methods and Evaluation

Making a genome sequence accessible and useful involves three basic steps: genome assembly, structural annotation, and functional annotation. The quality of data generated at each step influences the accuracy of inferences that can be made, with high-quality analyses producing better datasets that result in stronger hypotheses for downstream experiments. Here we report a new, high-confidence functional annotation of genes for the maize B73 reference genome. To develop this annotation set, we used sequence similarity- and protein domain-based methods as well as mixed methods developed for the Critical Assessment of Function Annotation (CAFA) competition. Individual annotation sets as well as combined outputs of multiple methods were compared both to each other and to the existing Gramene annotation (release 49) derived from the Ensembl pipeline. Our new functional annotation increases the number of genes that are assigned at least one functional annotation (GO term) as well as the quality of functional assignments on average (based on F-measure). Annotations derived from the GO Annotation Methods and Evaluation (GAME) pipeline will be made accessible via MaizeGDB (http://www.maizegdb.org) and figshare (https://figshare.com/projects/Mazie-GAMER/17150).