In silico prediction of active RNA genes in legumes

Accumulating evidence suggests that non-coding RNAs (ncRNAs) play key roles in gene regulation and may form the basis of an inter–gene communication system. MicroRNAs are a class of small non-coding RNAs found in both plants and animals that regulate the expression of other genes. Identification and analysis of microRNAs enhances our understanding of the important roles that microRNAs play in this complex regulatory network. The work presented in this thesis constitutes the first large-scale prediction and characterization of both ncRNAs and miRNAs in the model legume Medicago truncatula and Lotus japonicus, and provides a basis for further research on elucidating ncRNA function in legume genomics. Thus far the search for novel ncRNA genes is hampered by a lack of statistically significant sequence features for predicting ncRNA genes. However, many ncRNAs are synthesized similar to mRNAs and can be detected through screening of polyA-rich EST or cDNA libraries. In this thesis, I developed a computational pipeline to screen EST and genomic sequence data for those transcribed genes with limited protein coding potential and applied this pipeline to M. truncatula and L. japonicus. This process identified sets of 673 M. truncatula and 1637 L. japonicus mRNA-like RNA transcripts that appear not to encode proteins. Further computational analysis showed that many of these ncRNA candidates only had discernable homologues in closely related plants. By using a machine