A computational model for non-conserved mature miRNAs from the rice genome

Several computational approaches employ the high complementarity of plant miRNAs to target mRNAs as a filter to recognize miRNA. Numerous non-conserved miRNAs are known with more recent evolutionary origin as a result of target gene duplication events. We present here a computational model with knowledge inputs from reported non-conserved mature miRNAs of Oryza sativa (rice). Sequence- and structure-based approaches were used to retrieve miRNA features based on rice Argonaute protein and develop a multiple linear regression (MLR) model (r(2) = 0.996, q(2)cv = 0.989) which scored mature miRNAs as predicted by the MaturePred program. The model was validated by scoring test set (q(2) = 0.990) and computationally predicted mature miRNAs as external test set (q(2)test = 0.895). This strategy successfully enhanced the confidence of retrieving most probable non-conserved miRNAs from the rice genome. We anticipate that this computational model would recognize unknown non-conserved miRNA candidates and nurture the current mechanistic understanding of miRNA sorting to unveil the role of non-conserved miRNAs in gene silencing.