[b]Identification of Missing Genes in Rice Annotation by Histone Modifications[/b]   After discovering the high concurrence frequency between histone modifications and transcript regions, it was proposed that these marks could be used to identify unannotated genes. There are two widely used resources for gene annotation in rice (Japonica): MSU/TIGR (Ouyang et al., 2007) and RAP–DB (Itoh et al., 2007). We found 6629 RAP–DB genes whose TSSs did not overlap with any TSS region (500 bp up or downstream) in the MSU/TIGR gene, meaning that these genes were missing in the annotation. Among these unannotated genes, 471 genes had both H3K4me3 and H3K9ac peaks within 400 bp regions downstream of the TSS (Supplemental File 2). These results suggested that these genes existed but were missing in the rice MSU/TIGR annotation. To evaluate this prediction, two RAP–DB genes (Os02g0510400 and Os11g0564800), neither of which had a functional annotation, were selected from the list for experimental validation. Analysis using histone modifications showed that Os02g0510400 was located near a MSU/TIGR gene, LOC_Os02g30640, but had no overlap with any other gene (Figure 6A). The PCR experiments proved that Os02g0510400 existed and was transcribed under two different conditions: normal conditions and when treated with methyl viologen (MV) (Figure 6B). In another example, the Os11g0564800 overlapped with the MSU/TIGR annotated gene, LOC_Os11g35840, but was located on the antisense strand (Figure 6C). Intriguingly, both genes were transcribed, but under different conditions (Figure  6D). Os11g0564800 was transcribed after MV treatment and repressed under normal conditions, but LOC_Os11g35840 exhibited the opposite expression pattern. The results demonstrated that histone modifications that have a high concurrence possibility with active transcript loci could be used to predict unannotated genes—a valuable approach for poorly annotated organisms.