| Background. MicroRNAs (miRNAs) are a class of small noncoding RNAs that regulate the expression of other genes. Several computational methods were developed to identify miRNA genes in Arabidopsis thaliana. However, none of these methods can identify conserved and non-conserved miRNAs while maintaining high sensitivity and specificity at the same time. In addition, the number of miRNA genes in plants remains unclear.;Results. We present a genome-wide computational approach (implemented in a set of scripts termed "miRscore") to identify both conserved and non-conserved miRNA genes in Arabidopsis. When combined with experimental data from small RNA massively parallel signature sequencing and Northern blotting, our method captured 96 of 117 miRNA genes in version 8.0 of miRBase, including six genes not conserved in rice. In addition, we validated four new miRNA genes that are not conserved in rice, Populus, or Medicago. One of the predicted targets was validated by 5' RNA ligase-mediated rapid amplification of cDNA ends. We estimate that the maximal number of Arabidopsis miRNA genes conserved in rice is around 137, while the number of Arabidopsis-specific miRNA genes is more variable and 8 may be up to several hundred. The miRscore web server is available at http://www.plantgdb.org/prj/miRscore .;Conclusions. Features of miRscore include (1) high sensitivity and high specificity; (2) identification of both conserved and non-conserved miRNAs; (3) estimation of conserved and non-conserved miRNA gene numbers; and (4) publicly available web server. None of the previous methods can offer all these four features in a single solution. With all these advantages, miRscore can facilitate the discovery of additional conserved and non-conserved miRNAs in plants. |
