We present here a flexible computational framework to complete the large-scale computing tasks involved in automatic annotation of whole-genomes. The characteristics of this framework include a two-level job load system and NFS-based distributed store of replicated data. In addition, the storage structure of annotation results in a relational database system and a web interface for graphical interactive browsing and searching on the data are also described. The framework has been used to identify a core set of human protein coding genes that are consistently annotated and of high quality, which can be accessed by the browser provided at http://bioinfo.hust.edu.cn/en/database/ben.