Maize - GO Annotation Methods Evaluation and Review (Maize-GAMER)

Maize is an important agricultural crop - based on metric tons, maize is the #1 production grain crop in the world (http://faostat.fao.org). Largely as a result of plummeting sequencing costs, the growing availability of whole genome sequences has led to an increased demand for high-throughput omics studies. Interpretation of the results of these studies often requires the analysis of functional annotations of genes. The Gene Ontology (GO) is a widely used database that consists of terms that describe gene product function and property. The majority (~99%) of the GO annotations in maize are inferred from electronic annotations by high-throughput pipelines such as Ensembl.  On the contrary, only about half of the GO annotations in Arabidopsis are inferred from electronic annotations. Clearly there is a need to evaluate the confidence of existing GO annotations for maize and improve the overall quality of functional predictions in this well-studied species. Here we present a pipeline that annotates GO terms to maize gene models using multiple functional annotation methods along with an evaluation of confidence for these annotations. Our pipeline uses three approaches to assign GO terms: BLAST-based methods, functional domain-based methods, and advanced methods (machine learning and statistical approaches). Using a test dataset that contains high-quality manual annotations from MaizeGDB (~750) and reviewed annotations from UniProt (~6500), we are in the process of evaluating the performance of our approaches, comparing the performance to existing annotations, and creating a designated set of high-confidence functional annotations for maize genes. Review of these annotations by experts will substantially improve the confidence of annotations predicted using the pipeline, so we are designing a user-friendly system to leverage crowdsourcing for manual review of the predicted GO annotations. We plan to expand and adapt the pipeline to other species once it has been implemented and tested in maize.