If you measure a genome by the practical benefits of its sequencing, then they come no more important than that of rice (Oryza sativa) — the staple crop of over half the population. Moreover, it's a good model for other grasses, including the other global food crops wheat and maize.

Two draft sequences of rice are now reported in Science, both generated by whole-genome shotgun approaches. Jin Yu and colleagues, at the Beijing Genomics Institute (BGI), have sequenced the subspecies O. s. indica, which is commonly cultivated in China and South-east Asia, whereas Stephen Goff and colleagues at Syngenta have sequenced the O. s. japonica subspecies. Controversially, the BGI sequence is available in GenBank, whereas Syngenta's can only be accessed through an agreement similar to that used by Celera for its sequence of the human genome.

BGI have sequenced 362 Mb of the 466 Mb indica genome to fourfold coverage, ∼90% at over 99.9% accuracy. The initial annotation turned up 53,398 complete genes, although this could rise to 64,500 if many partial predictions are confirmed. The japonica genome is slightly shorter at 420 Mb, 390 Mb of which the Syngenta team sequenced at over sixfold coverage, at 99.8% accuracy. Their conservative annotation identified 32,277 genes, although the total number might be over 60,000.

Where the BGI group was able to align their draft to portions of the japonica genome sequenced by the International Rice Genome Sequencing Project, sequence identity was close to 100%. The 14% that did not match bears the hallmarks of transposon activity — evidence of rapid changes of genome size in the grasses.

Comparisons were also made with the Arabidopsis thaliana genome. Eighty-five per cent of Arabidopsis genes have homologues in rice, although only ∼50% of rice genes have homologues in Arabidopsis, hinting at a major duplication event in rice's evolutionary history. Around one-third of the homologous genes are plant specific. Also, ∼30% of the most highly conserved genes between the two plants have no confirmed function, showing that many crucially important plant genes remain to be characterized.

Deeper questions about the architecture and evolution of the rice genome await the finished sequence. Such 'gold standard' approaches are also required if the rice genome is to be used as a model for other grass staples, such as wheat, the large genome and hexaploid status of which make sequencing impractical.