Abstract
Sexual reproduction in plants underpins global food production and evolution. It is a complex process, requiring intricate signalling pathways integrating a multitude of internal and external cues. However, key players and especially non-coding genes controlling plant sexual reproduction remain elusive. We report the development of MCRiceRepGP a novel machine learning framework, which integrates genomic, transcriptomic, homology and available phenotypic evidence and employs multi-criteria decision analysis and machine learning to predict coding and non-coding genes involved in rice sexual reproduction.
The rice genome was re-annotated using deep sequencing transcriptomic data from reproduction-associated tissues/cell types identifying novel putative protein coding genes, transcript isoforms and long intergenic non-coding RNAs (lincRNAs). MCRiceRepGP was used for genome-wide discovery of sexual reproduction associated genes in rice; 2,275 protein-coding and 748 lincRNA genes were predicted to be involved in sexual reproduction. The annotation performed and the genes identified, especially the ones for which mutant lines with phenotypes are available provide a valuable resource. The analysis of genes identified gives insights into the genetic architecture of plant sexual reproduction. MCRiceRepGP can be used in combination with other genome-wide studies, like GWAS, giving more confidence that the genes identified are associated with the biological process of interest. As more data, especially about mutant plant phenotypes will become available, the power of MCRiceRepGP with grow providing researchers with a tool to identify candidate genes for future experiments. MCRiceRepGP is available as a web application (http://mcgplannotator.com/MCRiceRepGP/)
Significance statement Rice is a staple food crop plant for over half of the world’s population and sexual reproduction resulting in grain formation is a key process underpinning global food security. Despite considerable research efforts, much remains to be learned about the molecular mechanisms involved in rice sexual reproduction. We have developed MCRiceRepGP, a novel framework which allows prediction of sexual reproduction associated genes using multi-omics data, multicriteria decision analysis and machine learning. The genes identified and the methodology developed will become a significant resource for the plant research community.
