摘要：There are currently over 100 partially or entirely sequenced plant genomes but despite progress in sequencing, the function of the majority of plant genes remains largely unknown. Experimental researches cannot cover large-scale sequencing projects and this is where function prediction methods can assist. Most of them are based on sequence similarities but such approach is not entirely reliable because similarity does not always reflect homology. This study combines a phylogenomic approach with information from functional annotations, embedded into plant phylogenetic trees, which represent both sequence similarity and homology relationships between genes. Since functional annotations come from both manual and automatic annotations, an in-house semantic network approach was applied to improve their quality. This approach simulates the process of manual annotation and prevents generation of incorrect knowledge about gene function and domain distribution in de novo sequenced genomes. Then, the information from functional annotations, homology relationships and evolutionary distances based on tree topology was summarized for each gene using a scoring system for all potential functions. Functions with the highest score were assigned to uncharacterized genes or used as feedback to address inaccuracies in automatically annotated genes. As a result, genes were assigned into four major groups according to the score and the origin of their predicted function. These groups also reflect the reliability of the prediction. The predicted functional information is being involved in studying evolutionary relationships between genes responsible for the regulation of C3/C4 photosynthesis and can be applied to other major plant phenotypes.
关键词：functional annotation; function prediction; phylogenomics; plants; semantic network
