Noncoding RNAs (ncRNAs) are functional transcripts that do not encode proteins; they are involved in many regulation pathways. The only common feature shared by most (but not all) known ncRNAs is their ability to fold into secondary structures that are crucial for their functioning and thus should be conserved. This fact can be used for genome-wide prediction of ncRNAs. The approach employed in this study was based on computing local base pairing probabilities and further maximizing the total pairing probability for a segment using the Nussinoff-like algorithm. The suggested method was shown to efficiently predict known ncRNAs and possibly some new ncRNAs.