Sequence analysis of the Methanococcus jannaschii genome and the prediction of protein function

With the completion of the genome sequence of Methanococcus jannaschii (Bult et al., 1996), computational analysis has revealed a number of interesting predicted functions for this organism. Although the success rate of the initial prediction was <40%, due to a conservative attitude towards possible over-interpretation (Venter, 1996), additional efforts to annotate the sequence follow, contributing a significant increase of functional assignments through a combination of different methods (Kyrpides et al., 1996). In our continuing effort to annotate the gene products for each complete genome (Casari et al., 1995), we have analyzed the full genomic sequence of M.jannaschii, and predicted gene function by sequence similarity. We use GeneQuiz, a system for large-scale sequence analysis (Scharf et al., 1994), which exploits the combination of a number of predictive methods with a rule-based engine that increases the success of predictions thanks to a collection of heuristics and a number of benchmarking cycles (Casari et al., 1996). We have maintained a strong interest in the study of Archaea (Ouzounis and Sander, 1992; Ouzounis et al., 1995), and the analysis of the M.jannaschii genome was greatly anticipated. The scope of this communication is 3-fold: (i) to compare the performance of the GeneQuiz system against the laboriously derived (but highly accurate) manual annotation of the previous attempts; (ii) to discuss some of the cases where GeneQuiz has succeeded or failed; (iii) to assure potential users of the results that the quality of the analysis is high, despite the peculiar biochemical and phylogenetic disposition of this organism. From a total of 1682 chromosomal open reading frames (ORFs), GeneQuiz has assigned function to 774 (46%) with high confidence ('clear' cases), and, with decreasing confidence, 118 (7%) with probable function ('tentative' cases), while 482 (29%) ORFs have a clear homolog whose function remains unknown. The remaining 308 cases (18%)